LLM AI Startup Jobs

AI research company creating safe AGI that benefits all of humanity, known for ChatGPT and GPT models.

AI safety company building reliable, interpretable, and steerable AI systems, creator of Claude.

Cohere builds foundational AI models and enterprise solutions for search, generation, and agentic workflows. Their platform includes Command, Embed, Rerank, and the North agentic AI suite, enabling businesses to deploy secure, production-grade AI on private or sovereign infrastructure.

Cyera is an AI-native data security platform that helps enterprises discover, classify, govern, and protect sensitive data across cloud, SaaS, on-premises, and AI environments. The platform uses machine learning and LLMs for high-precision data classification at scale.

LangChain provides open-source frameworks and a managed platform (LangSmith, LangGraph) for building, observing, and deploying LLM-powered applications and agents. Its composable, modular components let developers integrate large language models into production workflows. Backed by Sequoia and IVP, LangChain is the leading developer framework for agentic AI.

Building application-specific AI chips (Sohu) optimized for transformer inference with 10x throughput gains.

Profound is the leading AI search visibility and optimization platform that helps enterprise brands understand and optimize how they appear across AI-powered search engines like ChatGPT, Perplexity, and other generative AI platforms. Serving over 10% of the Fortune 500 including Target, Walmart, and Figma, the company has pioneered the category of Answer Engine Optimization (AEO).

HappyRobot is an AI-native operating system for supply chain enterprises that deploys voice AI workers to handle calls, emails, and routine communications at scale. The platform combines real-time operational data, specialized AI agents, and orchestrating intelligence to automate mission-critical workflows. Founded in 2022 and backed by a16z and Base10 Partners, it serves over 70 enterprise logistics customers.

Perplexity builds an AI-native answer engine and assistant that combines web retrieval with LLMs for consumer and enterprise search workflows.

AI care partner platform for clinicians that automates clinical documentation, streamlines workflows, and improves patient care. Used across Australia, the US, UK, and Canada with partnerships including the NHS.

Cognition is an AI agent lab building software engineering agents, including Devin, to help engineering teams plan, implement, and debug software more autonomously.

Hippocratic AI builds safety-focused healthcare LLMs and voice agents for non-diagnostic patient engagement, navigation, education, and administrative workflows.

Perle is an AI training data platform that combines human expertise with adaptive workflows to help companies collect, annotate, and evaluate specialized training data for generative AI, LLMs, and RLHF. Their vetted global network of domain experts provides modular solutions for data annotation, enrichment, and adversarial robustness assessment.

Together AI provides an AI-native cloud platform for inference, fine-tuning, GPU clusters, and large-scale model training, helping AI engineers and researchers build and scale open-source and enterprise AI systems.

Reflection AI is building frontier open intelligence accessible to all, developing open-weight AI models. Founded by former Google DeepMind researchers who co-created AlphaGo, raised $2B at an $8B valuation.

Secure generative AI platform purpose-built for elite financial institutions, automating investment banking research and analysis. Founded by former investment banking analysts, Rogo serves major banks with AI-powered financial agents across offices in New York and London.

Suno is an AI music generation platform that enables anyone to create original songs from text prompts. Based in Cambridge, MA, it has reached $200M in annual revenue and is one of the fastest-growing generative AI startups.

WRITER provides a full-stack enterprise generative AI platform for building, activating, and supervising AI agents and workflows grounded in company data, with enterprise-grade LLMs, governance, and deployment tools.

Snorkel AI provides a data-centric AI platform for building specialized AI systems, helping enterprises and frontier teams create, curate, evaluate, and govern expert data for production models and agents.

Builds digital in-memory compute hardware and software stack for low-latency, energy-efficient LLM inference.

Emergent is an AI app-building platform focused on no-code and agent-assisted product creation, helping users rapidly build real applications with AI.

GC AI builds AI software for in-house legal teams, combining contract review, playbooks, research, and workflow automation in a secure legal workspace.

Arize builds AI observability and evaluation software for machine learning, LLM, and agent systems, helping teams trace, monitor, and improve production model performance.

Assort Health builds agentic AI voice agents for healthcare call centers, automating appointment management and patient phone calls with specialty-specific generative AI. The platform handles patient access workflows that previously consumed thousands of human hours, improving patient experience while reducing operational costs. With $102 million in total funding, Assort Health serves major healthcare organizations across the United States.

AI research lab founded by former OpenAI CTO Mira Murati, building Tinker, an API for distributed LLM fine-tuning that makes model customization accessible without massive GPU infrastructure.

Factory is the platform for agent-native software development, building AI-powered coding agents called Droids that automate tasks across the entire software development lifecycle. Used by enterprise engineering teams at MongoDB, EY, and Zapier, Factory delivers up to 31x faster feature delivery and 96% shorter migration times.

Serverless cloud platform for developing, deploying, and scaling AI applications. Developers can run any code in the cloud without managing infrastructure, with built-in support for GPU workloads and batch jobs.

Reducto provides a high-quality AI document ingestion and parsing API for large language models. The Y Combinator-backed company processes nearly a billion pages monthly for leading AI teams like Harvey and Scale AI.

AfterQuery is a San Francisco-based applied research lab that builds expert-curated AI training datasets and model evaluation suites for frontier AI labs. It leverages a network of nearly 100,000 domain experts (developers, attorneys, and other professionals) to generate high-quality data, evals, and dev environments. The company reached over $100M ARR within roughly 14 months of founding.

Bluefish (Bluefish Labs) is an AI marketing platform for enterprise brands that helps Fortune 500 companies gain visibility and influence over how their products surface in AI-generated answers and product discovery. The platform analyzes millions of prompt responses across major AI systems to deliver insights and optimizations that improve brand presence, consistency, and favorability. Customers include Adidas and Tishman Speyer.

H Company is a frontier AI company building autonomous agents and foundation-model systems for enterprise work. The company develops research, model, inference, and infrastructure capabilities behind products like Runner H.

Viktor is an AI virtual coworker embedded inside Slack and Microsoft Teams that connects to thousands of workplace tools to answer questions and execute multi-step workflows. It reached a $15M ARR run rate within three months of its February 2026 public launch, serving more than 2,000 organizations. In May 2026 it raised a $75M Series A led by Accel.

Spellbook builds an AI copilot for transactional lawyers, helping legal teams review, draft, and negotiate contracts directly inside their document workflows.

Prime Intellect is an AI infrastructure company that enables training and deploying frontier AI models across decentralized, globally distributed compute resources, commoditizing both compute and intelligence for researchers and organizations.

Goodfire is an AI interpretability research company focused on understanding and designing AI systems through its flagship platform Ember, which decodes the neurons inside AI models to give direct, programmable access to their internal thoughts and decision-making processes.

Sesame AI builds lifelike voice companions and AI interfaces with advanced conversational speech models. Their flagship AI model Maya powers a personal assistant, and they develop lightweight eyewear hardware with integrated audio.

AI-powered legal search and due diligence platform that enables law firms and legal teams to instantly search, analyze, and surface insights across large document repositories.

InstaLILY AI builds domain-trained AI agents ("InstaWorkers") that operate inside existing enterprise systems like ERPs, CRMs, and email to autonomously execute sales, service, and operations workflows. It targets distribution-heavy and regulated industries such as construction supply, OEM equipment, insurance, and healthcare, with optional human-in-the-loop oversight. The company raised a $25M Series A in August 2025 led by Insight Partners.

Robotics AI company building foundation models for general-purpose robots to handle physical world interactions with advanced manipulation capabilities

Braintrust builds evaluation and observability infrastructure for AI applications, enabling teams to test prompts, monitor quality, and ship LLM features with higher reliability.

Hebbia is an AI platform purpose-built for institutional financial services, enabling asset managers, investment bankers, and Fortune 500 companies to analyze vast quantities of financial documents and data with precision AI agents designed for billion-dollar decisions.

Latent is a San Francisco-based clinical AI startup automating medication access for health systems. Its medical language models read EHR patient data, interpret drug approval criteria, and orchestrate prior authorization, 340B compliance, and appeals workflows to close the gap between diagnosis and treatment. It partners with major US health systems including Yale New Haven, Mount Sinai, Vanderbilt, UCSF, and UCLA.

Periodic Labs is an AI research company focused on building autonomous AI scientists and software systems that accelerate scientific discovery.

Wordsmith builds legal AI software that helps in-house legal teams review contracts, draft documents, and automate legal workflows with LLM-powered tools.

Inception (Inception Labs) is a Palo Alto-based AI research company building diffusion large language models (dLLMs) that generate text and code in parallel, delivering up to 10x faster inference and lower cost than traditional autoregressive LLMs. Its Mercury model family targets latency-sensitive, real-time applications. Founded in 2024 by Stanford professor Stefano Ermon with Aditya Grover and Volodymyr Kuleshov.

Parallel builds web infrastructure for AI agents, providing APIs and tooling that let models search, browse, and interact with the live web more reliably.

Generalist AI is building embodied foundation models for general-purpose robots, starting with a focus on dexterous manipulation, leveraging research from OpenAI, Boston Dynamics, and Google DeepMind to create end-to-end deep neural networks that map sensor data to real-time robotic actions.

Krea is a next-generation AI creative suite that provides a unified platform combining multiple generative AI models with a custom interface for image, video, and multimedia creation. It is designed to give designers and creatives intuitive, fine-grained control over AI-generated content. Customers include Pixar, LEGO, and Samsung.

Finch is an AI-powered pre-litigation platform for personal injury law firms, pairing experienced paralegals and case staff with AI agents to automate intake, case management, medical records, lien management, and demand drafting. Founded in 2025 and based in New York City, it launched with a Sequoia-led seed round and raised a $20M Series A led by Redpoint Ventures.

Foundation model company building efficient general-purpose AI systems and Liquid Foundation Models for edge and resource-constrained environments.

LMArena develops and operates large-model evaluation infrastructure, including comparative benchmarks and community-driven testing workflows for frontier LLMs.

AI-first protein design company using deep generative models to create novel, functional proteins for applications in biomedicine, agriculture, and gene editing.

Flip CX (formerly RedRoute) is a vertical Voice AI company that automates inbound customer support phone calls for retail/eCommerce, healthcare, and transportation businesses. Its Alexa-like voice assistant resolves up to 90% of routine inbound calls and has automated 300M+ calls for 250+ enterprise brands including Under Armour, Tory Burch, and Newell Brands. The company reached eight-figure ARR growing 3x year over year.

Langdock builds a secure enterprise AI workspace for deploying LLM assistants, enterprise search, and internal workflows across company knowledge and tools.

OpenRouter provides developers with seamless access to 500+ AI models through a single unified API, eliminating the need to rewrite code or renegotiate contracts for LLM integration.

Neuromorphic computing startup building brain-inspired analog hardware to achieve biology-scale energy efficiency for AI workloads, founded by former Databricks AI head Naveen Rao.

Blackstar is an AI hardware startup building a new computing device and operating system aimed at fundamentally improving how humans interact with AI. Founded by former OpenAI Codex engineer Daniel Edrisian, the ~8-person team is split between San Francisco and Shenzhen and has not yet shipped a public product.

Cogent Security is the first company to deliver an AI taskforce dedicated to vulnerability management, using autonomous AI agents to handle the full lifecycle from threat identification to remediation. The platform autonomously prioritizes critical risks and executes fixes with minimal human oversight, serving dozens of Fortune 1000 companies. Founded in 2024 and incubated by Greylock Partners, Cogent has raised $53 million to date.

Granola builds an AI notepad for meetings that records device audio without a bot, turns rough notes and transcripts into structured meeting notes, and helps teams reuse meeting context in follow-up work.

LlamaIndex is a data framework for LLM applications that enables developers to connect, index, and query custom data sources with large language models through their open-source library and LlamaCloud platform.

Nace AI builds MetaModel, an AI system that dynamically generates small, task-specific language models for enterprises using a Hypernet architecture to inject task-specific weights in real-time without retraining. Their first product NAVI applies this technology to audit, compliance, and finance workflows.

Norm AI is a legal and compliance AI platform that transforms regulations into autonomous AI agents capable of monitoring policies and evaluating content against regulatory requirements. The company serves compliance officers at Fortune 100 companies and institutions managing over $30 trillion in assets, offering solutions for contract review, regulated content review, and compliance automation. It operates a proprietary Legal Engineering Automation Platform (LEAP) built by attorneys and former regulators.

AI code generation platform using reinforcement learning to build enterprise software development tools.

Wonderful is an enterprise AI platform that lets companies build, test, deploy, and monitor AI agents handling customer and employee interactions across voice, chat, email, and back-office systems in any language. It pairs the agent platform with locally embedded, forward-deployed teams to drive pilots into production. Founded in 2025 by Bar Winkler (ex-Approve.com) and Roey Lalazar, it scaled to a $2B valuation within roughly a year.

AI research company building adaptive models that learn continuously without expensive retraining, using less computing power than leading AI models, founded by former Cohere VP of Research Sara Hooker

Legion Intelligence (formerly Yurts) builds governed AI agent-orchestration software for defense, public sector, and enterprise operations. Its platform lets organizations securely deploy, manage, and scale autonomous AI agents across sensitive networks (including IL-4/5/6) for decision support and operational intelligence, with deployments at U.S. Special Operations Command and other DoD environments.

Lightfield is an AI-native CRM that automatically captures and understands customer interactions, eliminating manual data entry and note-taking for sales teams. Built by the founders of Tome (the presentation tool with 25 million users), Lightfield provides complete customer memory that updates after every meeting and conversation. The platform enables founders and salespeople to focus on closing deals rather than secretarial work.

Thread AI builds enterprise AI infrastructure with a knowledge-graph-based context engine to improve AI application accuracy and reliability.

Ada is an AI-powered customer service automation company that enables enterprises to resolve customer inquiries at scale using AI agents. The platform handles billions of interactions across 50+ languages for brands like Meta, Verizon, and Square.

August is the first AI platform built specifically for midsize law firms, enabling configurable legal AI workflows for contract review, due diligence, discovery preparation, and drafting. Founded by a team from Columbia University's ML research lab, August serves an underserved market of over 50,000 midsize law practices worldwide employing half a million lawyers.

Patlytics is an AI-native platform built for the entire patent lifecycle — from application drafting and prosecution to infringement analysis, invalidity research, and portfolio management. Trusted by over 40% of the Am Law 100 and corporate IP teams, it helps patent professionals do higher-quality work faster. Headquartered in New York City with offices in London and Toronto.

SF Compute provides rentable, large, low-cost GPU clusters for AI pre-training workloads. The platform operates as a marketplace connecting AI teams with on-demand high-performance computing capacity, offering flexible access to supercomputing-scale infrastructure with InfiniBand interconnects.

Titan is a banking-native AI platform purpose-built for regulated financial institutions. Its proprietary small language models are trained on OCC, CFPB, FFIEC, FDIC, and Federal Reserve guidance and deployed as governed agents that help bankers draft credit memos, summarize loan files, extract regulatory guidance, and automate compliance and lending workflows. Every response ships with a chain-of-thought audit log for auditability.

AirOps builds content engineering software that helps brands research, generate, and optimize content to win AI search and organic growth.

Applied Compute is a San Francisco-based AI startup founded by former OpenAI researchers that builds 'Specific Intelligence' — proprietary AI agents trained on a company's own operations and data that improve with every interaction. It develops custom AI models and a forward-deployed agent platform for enterprises. The company launched publicly in October 2025 and reached a $1.3B valuation in April 2026.

Bespoke Labs is an applied AI company that builds the reinforcement learning environments, data curation systems, and infrastructure used to train, evaluate, and post-train reliable long-horizon AI agents for frontier labs and enterprises. It is known for open-source tooling like Bespoke Curator and open datasets. Founded in 2024 by Mahesh Sathiamoorthy and Alex Dimakis, it is headquartered in Mountain View, California.

F2 is an AI platform for private markets investors, purpose-built for private credit, private equity, and commercial banking teams. It consolidates deal information scattered across financial models, CIMs, and memos into a unified, queryable format to speed up screening, underwriting, and portfolio monitoring—helping teams evaluate deals over 70% faster. Spun out of Arc Technologies in 2025, it is headquartered in New York City.

AI-powered medical information platform that synthesizes peer-reviewed clinical literature for physicians, functioning as an always-on clinical decision support tool used by hundreds of thousands of doctors.

Ankar AI builds an AI operating system for patent teams, supporting drafting, office-action workflows, and patent intelligence with legal-domain AI tools.

AI document generation platform for life sciences that automates regulatory paperwork for diagnostic, medical device, and drug development companies. Founded by Surbhi Sarna and Nate Smith.

GMI Cloud is an AI-native GPU cloud platform offering serverless inference and dedicated NVIDIA GPU clusters (H100, H200, and Blackwell) for training and running production AI workloads. Founded in 2022 and headquartered in San Jose, CA, the company serves startups and enterprises seeking scalable GPU infrastructure as an NVIDIA Preferred Partner.

Rowspace is a San Francisco-based AI platform that helps financial services firms—especially private equity and investment firms—turn proprietary, scattered internal data into actionable decision intelligence. It unifies information across document repositories, investment platforms, and accounting systems to support investment research, risk analysis, and deal evaluation. Founded in 2024 by former Notion CTO Michael Manapat and CFO Yibo Ling, it emerged from stealth in February 2026.

You.com builds AI search and retrieval infrastructure for teams that need domain-specific, production-ready search inside their own products and workflows.

CrewAI builds a multi-agent platform and open-source framework for orchestrating collaborative AI agent workflows in production applications.

Evertune is a Generative Engine Optimization (GEO) and AI marketing platform that helps enterprise brands measure and improve their visibility in AI search and LLM-generated answers. Founded by Trade Desk veterans, it analyzes how large language models select and rank brands, then delivers data-driven insights and content strategies to increase AI discoverability.

Heron Data uses AI and LLMs to automate document-heavy workflows in lending, equipment finance, and insurance, extracting and enriching information from documents at scale. The company processes over 350,000 documents per week for 150+ customers including FDIC-insured banks and insurance carriers. It closed a $16M Series A led by Insight Partners in July 2025.

Patronus AI builds evaluation and reliability tooling for enterprise AI systems, helping teams test, monitor, and improve LLM behavior in production.

Pulse is a production-grade unstructured document extraction platform that converts complex PDFs, Word, Excel, and other document formats into structured, LLM-ready data. Built for enterprise reliability with OCR, bounding boxes, and vision language model capabilities.

Tolans (Portola) is developing an AI-powered conversational companion in the form of playful alien characters designed to enhance creativity, imagination, and emotional connection, offering every person on Earth a personalized embodied AI best friend.

Open-source data preprocessing platform that extracts, cleans, and transforms unstructured documents (PDFs, images, HTML, emails) into structured formats optimized for AI and LLM pipelines.

WorkWeave (also branded 'Weave') is an AI-powered engineering analytics platform that quantifies software engineering output by analyzing every pull request and developer activity to measure real work and the impact of AI coding tools. It gives engineering leaders visibility into team performance and AI investment ROI. Founded by the ex-Causal team and backed by Y Combinator (W25), it serves 150+ customers including Rho, 11x, and Browserbase.

Mechanical Orchard uses AI to modernize legacy enterprise software and mainframe systems, helping large organizations transform older codebases while preserving critical business logic. Its product, Imogen, captures actual system behavior to guide incremental rewrites into modern cloud-native code. Founded by Rob Mee (ex-CEO of Pivotal Labs), it targets enterprises seeking to reduce technical debt without disrupting operations.

Physical AI company building Newton foundation model that fuses real-time multimodal sensor data with natural language to interpret and act on physical phenomena

Platform for creating a digital you—an AI persona trained on your content and style for Q&A and interactions.

Endex builds Excel-native AI agents for financial modeling and analysis, helping finance teams automate spreadsheet-heavy workflows.

Fiddler AI provides an AI observability and governance platform for monitoring model performance, drift, bias, and risk across machine learning and generative AI systems. It acts as a control plane for AI agents and models, helping enterprises ensure safety, transparency, and accountability in production. Customers span financial services, retail, and technology. The company was formerly known as Fiddler Labs.

Hugging Face is the leading open-source AI platform for building, training, and deploying machine learning models. It hosts hundreds of thousands of models and datasets, serving as the central hub for the AI developer community.

Inflection AI develops personal AI assistant products and foundation-model capabilities for natural language interaction and conversational intelligence.

Raindrop is an observability and monitoring platform for AI agents, often described as 'Sentry for AI products.' The platform helps engineering teams detect, track, and fix silent failures in AI agents running in production environments. Founded by Y Combinator alumni and backed by Lightspeed Venture Partners, Raindrop provides automated issue discovery, triage, and detailed failure explanations for AI systems.

Frontier AI research lab building next-generation AI models for semiconductor design. Develops recursive self-improvement cycle where AI designs silicon that powers the next generation of AI. Founded by AlphaChip creators from Google DeepMind.

Sekai (incorporated as Versa Inc) is a Sunnyvale, CA-based AI platform founded in 2024 by Lucky Zhang and Arthur Cheung, a founding member of TikTok. The platform lets users create, play, and remix AI-generated mini apps via text prompts, with over 15 million mini apps created and more than 200,000 generated daily.

StudyFetch is an AI-powered education platform that converts notes, PDFs, lectures, and videos into interactive flashcards, quizzes, and personalized study guides for students. Founded in 2023 and headquartered in Los Angeles, CA, its AI tutor and study tools help millions of students turn course materials into adaptive, self-paced learning sets. It is backed by Owl Ventures and College Board.

The Interaction Company of California builds Poke, a consumer AI agent accessible over messaging channels for automations, scheduling, and everyday task execution.

Arcade AI is an MCP runtime platform that enables AI agents to take secure, authenticated actions across business systems. The platform provides agent-optimized tool calling, OAuth-based authorization, and governance infrastructure for production AI deployments.

Didero is a US-based AI procurement and supply-chain startup that builds agentic AI sitting on top of a company's existing ERP to automate core procurement workflows for manufacturers and distributors—supplier communication, order tracking, exception handling, and spend analysis. Founded in 2023 and based in New York, it is embedded with 30+ customers. In February 2026 it raised a $30M Series A co-led by Chemistry and Headline with participation from Microsoft's M12.

Inherent is a London-based frontier AI lab building Faraday, an AI system for open-ended scientific discovery where humans and self-improving AI collaborate to tackle hard problems in science. Founded by ex-DeepMind researchers, it aims to build AI that recursively self-improves to explore 'unknown unknowns' and surface paradigm-shifting research questions.

Inkeep provides a platform for customer-facing AI agents and automations, combining no-code builders with developer SDKs for support, product, GTM, and operations teams.

Open-source AI company building tools to structure, understand, and collaborate with unstructured data including GPT4ALL, Atlas visualization platform, and Nomic Embed text embeddings.

Engineering general intelligence (“Archie”) that performs quantitative and spatial reasoning for physical engineering.

Raylu is an AI-native deal origination platform for private-market investors that automates market mapping, target research, and outreach. Venture, growth equity, private equity, and corporate development teams use it to go from investment thesis to booked founder meetings in minutes instead of weeks. More than 45 funds representing over $500B in AUM use Raylu as their AI sourcing engine.

Scaled Cognition is a reliability-focused AI lab building APT (Agentic Pretrained Transformer), a frontier model purpose-built for enterprise customer experience that aims to eliminate hallucinations and enforce policy compliance. It was founded in 2022 by CTO Dan Klein (a UC Berkeley NLP professor) and CEO Dan Roth, who previously built and sold an early agentic AI company (Semantic Machines) to Microsoft. The model is already embedded in partners like Genesys for agentic virtual agents across telecom, travel, banking, and retail.

Vectara provides a retrieval-augmented generation platform with enterprise search, grounded generation, and evaluation tooling to help teams build production-grade generative AI assistants.

alphaXiv is an AI research platform that helps users understand academic papers in minutes by bringing together papers, benchmarks, and implementations in one place. The platform bridges the gap between AI research and practice, enabling engineers to quickly surface relevant papers, compare methods, and interact with implementations. Since launching in 2024, alphaXiv has reached millions of users across academia and industry.

AMI Labs is a frontier AI research lab building world-model-based intelligence systems with planning, memory, and safety controls for real-world applications.

Blockit AI builds an autonomous scheduling agent that books, reschedules, and coordinates meetings over email and Slack, integrating with Google and Outlook calendars. It learns users' time preferences and negotiates directly with other people's calendar agents to eliminate scheduling back-and-forth.

Botpress is an AI agent infrastructure platform that gives developers the tooling to build, deploy, and scale production-ready AI agents. Founded in 2017 and headquartered in Montréal, the company raised a $25M Series B in June 2025 to expand its agent infrastructure stack.

Open-source AI coding agent that runs in your IDE to plan, write, and edit code with your project context.

Gradium is a Paris-based AI company building ultra-low latency voice foundation models that enable near-instant, expressive, and multilingual voice AI at scale. Spun out of French AI lab Kyutai, the company was founded by former Google DeepMind researchers who pioneered key advances in audio language modeling. Backed by a $70 million seed round from FirstMark Capital, Eurazeo, Eric Schmidt, and Xavier Niel, Gradium aims to make voice the universal interface for AI.

Graphon AI is a pre-model intelligence layer that transforms enterprise multimodal data (documents, video, audio, logs, databases) into persistent relational memory using graphon mathematical functions. It enables LLMs and AI agents to reason accurately across unlimited connected data sources with larger effective context. The company was founded by researchers from Amazon, Meta, MIT, Google, Apple, and NVIDIA.

Hume AI develops empathic AI models and APIs for voice, speech, and expression understanding, including the Empathic Voice Interface for real-time emotionally intelligent conversations.

Magic is an AI coding agent company building autonomous software engineers capable of handling complex, long-horizon coding tasks. Its system combines frontier LLMs with proprietary long-context retrieval and tool-use to write, test, and deploy production-ready code.

Paradigm is an AI-native spreadsheet platform that puts swarms of intelligent agents at your fingertips. Each cell can be powered by its own AI agent to crawl the web, fill in data, and automate structured data workflows. Backed by General Catalyst and Y Combinator.

Prox builds AI technical support agents for complex physical products such as power tools, HVAC systems, industrial heaters, EV chargers, and CNC machines. Its multimodal platform is grounded in product manuals and cited to the page, and can draw wiring schematics, share CAD models, and handle voice and video support across product pages, dealer counters, internal tools, ChatGPT, and Claude. The company serves hardware and industrial-equipment makers in powersports, agriculture, HVAC, firearms, and construction equipment.

Reka builds multimodal foundation models and enterprise AI systems that combine text, image, audio, and reasoning capabilities for production applications.

ZeroDrift is an AI compliance firewall that enforces regulatory and company-policy rules on AI-generated messages before delivery, blocking, rewriting, or approving them in real time across chatbots, voice, and AI agents. Deterministic checks flag violations against standards like SOC 2 and GDPR, then an LLM rewrites a compliant version of the message. It targets tier-one banks, asset managers, and insurers deploying AI in regulated communications.

Anara (formerly Unriddle) is an AI-powered research assistant that helps scientists and academics rapidly find insights across research papers, enabling faster literature reviews and scientific discovery.

Cleric builds an AI site reliability engineer that investigates incidents, reasons across production signals, and helps engineering teams resolve operational issues faster.

General Legal is an AI-native law firm focused on growth-stage companies, providing fast, high-quality contract drafting, review, and negotiation for a flat fee. Founded by former Casetext and Thomson Reuters alumni, it combines experienced attorneys with proprietary AI workflows to deliver 10x the efficiency of traditional law firms. Handles MSAs, NDAs, DPAs, and vendor agreements with guaranteed turnaround times.

Highlight AI (legal name Highlight Holding, Inc., also known as Embedded Intelligence) builds a shared intelligence layer that captures commitments, connects context across tools like Slack, Notion, GitHub, and Gmail, and automates next steps for teams. Its product auto-drafts tickets and documents from meetings, tracks commitments, and automates recurring workflows for the agentic age of work. Founded in 2024 and led by CEO Sergei Sorokin (former Discord VP of Product), it raised a $40M Series A led by Khosla Ventures in March 2026.

Inference provides a distributed, serverless platform that lets developers run open-source AI models without managing infrastructure. The company aggregates underutilized data center capacity to offer cost-effective LLM inference via a simple API.

Julius is an AI-powered data analysis platform that acts as a personal data scientist, enabling users to analyze and visualize datasets and perform predictive modeling through natural language prompts. With over 2 million users generating more than 10 million visualizations, Julius makes data science accessible to everyone.

Langbase is a serverless AI developer platform for building, deploying, and scaling AI agents. It processes 700TB+ of vector memory and billions of AI agent runs monthly, serving 100,000+ developers with support for 600+ LLMs.

Lindy is an AI agent platform that lets users build and deploy AI employees to automate repetitive knowledge work including inbox management, meeting scheduling, customer support, and lead research. Their platform features an Agent Builder for prompt-to-agent creation and Autopilot for browser and desktop automation tasks.

Luminos.AI builds automated governance and testing tools for generative and agentic AI systems, built by lawyers and data scientists. The platform helps legal and compliance teams assess AI risk, run compliance evaluations, and generate automated reporting to detect legal, regulatory, safety, and security liabilities before AI products are deployed.

Mainframe is building generative computing software that lets teams orchestrate AI agents and automate knowledge work through its Cobot product.

Mem0 is a universal, self-improving memory layer for LLM applications that enables them to remember and learn from user interactions over time, powering personalized AI experiences with over 41,000 GitHub stars and 14 million Python package downloads.

Frontier AI lab building reasoning models that let anyone describe a game or simulation in natural language and see it transformed into a working interactive 2D/3D scene within minutes.

Slingshot AI is a mental health research lab building a foundation model for psychology. Its flagship product Ash is an AI chatbot designed for therapy, providing 24/7 mental health support. Backed by a16z and Forerunner, the company has raised $123M to make mental healthcare more accessible.

Tomo is a personal AI 'champion' that lives inside your text messages: you text a phone number and it learns your goals, remembers what matters to you, and sends reminders and motivation to help you follow through. It can join group chats, track calories from food photos, budget from bank-statement screenshots, and connect to your calendar, email, Notion and Google Drive. The company emerged from stealth in June 2026 after reaching over 10,000 paying customers on word of mouth.

Variant is a creative intelligence platform that brings taste to code generation, allowing users to generate endless design variations for apps and websites from a single idea. Backed by Y Combinator (F24), the tool combines AI-powered design sensibility with code generation to help developers build better software through visual exploration.

Willow develops AI voice dictation software designed to dramatically speed up professional writing and communication workflows across email, documents, and messaging tools.

ZeroEntropy provides a high-accuracy search API over unstructured data for AI agents and RAG applications. The YC-backed company builds smarter retrieval models enabling AI agents across healthcare, law, and sales.

Browser Use is an open-source framework that enables AI agents to autonomously navigate and interact with websites like a human user. Built by ETH Zurich alumni, the project has amassed over 80,000 GitHub stars and powers agentic AI platforms.

Conductor is a developer-tools startup that lets engineers run a team of AI coding agents (Claude Code, Codex, Cursor) in parallel, isolated workspaces on their Mac, then review and merge the agents' work. Launched in 2025, it is used by engineers at companies like Linear, Vercel, Notion, and Stripe. It raised a $22M Series A led by Matrix and Spark in March 2026.

AI teammate that researches, plans, and executes multi-step work across 25+ enterprise apps using deep company context (OM1).

YC W24 AI platform that translates codebases into structured documentation, architecture maps, and history guides, compressing technical discovery from months to days.

Endeavor builds AI-powered agents for enterprise revenue operations, automating complex workflows across sales, finance, and customer success. Its platform integrates with existing enterprise systems to execute multi-step business processes autonomously.

Fyxer provides AI assistants for email and meetings, automating inbox management, scheduling, and routine coordination tasks for professionals.

Enterprise AI marketing platform that helps teams generate on-brand content and campaign assets at scale, with workflows and AI copilots for marketing operations.

Pythagora builds an AI software development platform that helps teams generate, test, and ship applications faster, building on its open-source GPT Pilot project.

Semantic KV caching layer built for LLM inference, enabling AI applications to reduce inference costs and latency by reusing cached computation across similar prompts.

Autonomous is an AI-powered financial advisor that pairs robo-advisor automation with the personalization of a human wealth manager at a 0% advisory fee. Built by Autonomous Technologies Group (ATG), an applied AI research lab founded by Paperspace co-founders Dillon Erb and Daniel Kobran, it uses frontier AI models and real-time market data to build and manage personalized investment portfolios. The consumer iOS app aims to be SOC2, FINRA-licensed, and SEC-compliant at launch.

Boson AI builds proprietary multimodal foundation models for voice and audio (Higgs) and Feynman Flow, an agentic platform that turns model capability into deployed intelligence. Founded in 2023 by former AWS AI leaders Alex Smola and Mu Li, the company focuses on audio/voice understanding, generation, and real-time interaction. The team of 30+ researchers and engineers operates across Santa Clara and Toronto.

AI-native operating system for biopharma companies, unifying internal and external data, codifying decision logic, and generating critical deliverables across the drug development lifecycle.

Deduction is a New York-based consumer fintech startup building Taylor, CPAI — an AI tax accountant for U.S. consumers that pairs an autonomous AI agent with licensed-CPA oversight. The platform delivers on-demand tax filings, year-round tax planning, and instant answers via call, text, chat, or email at a fraction of a traditional accountant's cost.

Builds games that generate reinforcement learning training data for frontier AI models, enabling millions of players to naturally evaluate and improve AI systems through gameplay.

Guru is an AI source-of-truth platform for knowledge work that connects company knowledge, delivers verified answers in workflow tools, and keeps information accurate automatically.

HoneyHive provides observability, evaluation, and monitoring tools for teams building production AI agents.

Humans& is a human-centric AI research lab focused on building agentic AI systems and frontier models.

Imbue (formerly Generally Intelligent) is an AI research lab building agents that can robustly reason and code. The team trains foundation models optimized for reasoning and ships applied tools (Sculptor, Bouncer) that let anyone build and control custom AI agents. The company has raised over $200M to support its long-term mission.

Lazarus AI (Lazarus Enterprises, Inc.) is a Boston/Cambridge, MA-based enterprise AI company focused on document understanding and intelligent data processing. Its platform — OCR/Forms extraction, the RikAI reasoning engine, and a Vector Knowledge Graph — extracts and reasons over documents, images, and other unstructured data without training, serving insurance, public-sector, and healthcare customers. It received undisclosed strategic investments from WaveFx (Aug 2024) and QBE Ventures (Oct 2024).

LM Studio is a desktop application and developer toolkit built by Element Labs Inc. that lets users discover, download, and run large language models locally and privately on their own hardware. It provides an OpenAI-compatible local API server and supports Windows, macOS, and Linux. Founded in 2022 by Yagil Burowski and headquartered in Brooklyn, New York.

Mastra is an open-source TypeScript framework for building AI agents and applications, created by the team behind Gatsby. It provides modular tools for agent orchestration, tool calling, and workflow automation.

Nous Research is an open-source AI lab developing large language models and decentralized training infrastructure, including Hermes models and the Psyche network.

Nuraline is an applied AI company building a forward-deployed AI agent that enables AI systems to continuously self-improve in production by connecting telemetry, traces, and user feedback to generate evaluations and tested improvements across the technology stack. The platform works closely with product teams and enterprises requiring measurable reliability gains in reasoning-heavy applications like enterprise search and web navigation. Currently operating as Introspection (introspection.dev), the company is in its earliest stage of development.

Orchids is an AI-powered app and website builder that turns simple prompts into full-stack applications without coding. Supports web apps, mobile apps, Slack bots, browser extensions, and AI agents. YC W25 batch.

Radical AI is building next-generation AI infrastructure with a focus on custom silicon and compute systems optimized for large-scale model training and inference. It aims to reduce the cost and energy consumption of AI workloads through purpose-built hardware and software co-design.

Decentralized open-source AGI platform backed by Founders Fund and Pantera Capital, building an open AGI network where multiple AI agents collaborate in real time to prevent single-entity AGI control.

dottxt builds structured generation tooling for production LLM applications, helping developers force reliable outputs that plug into existing software systems.

Augment Code is an AI coding assistant built for professional software engineers, offering deep codebase understanding through long-context retrieval and enterprise-grade security. It integrates with existing development workflows to accelerate code generation, review, and refactoring at scale.

Ceramic.ai builds enterprise AI training infrastructure that improves the speed, efficiency, and scalability of large-model development.

Dedalus Labs builds infrastructure for agentic AI, with products and APIs for deploying multimodal, multi-tool agents and experimenting with autonomous workflows in production.

Layercode provides voice AI infrastructure for developers, enabling production-ready, low-latency voice AI agents using TypeScript. The platform handles real-time audio processing across 330+ global edge locations with sub-50ms latency, supporting multiple LLM and voice providers.

Onlook is an open-source visual IDE for React and Next.js applications, described as 'Cursor for Designers.' It lets teams design directly in their real codebase by visually editing components while AI syncs changes back to source code in real time. Founded in 2024 and backed by Y Combinator (W25) and Leonis Capital.

Onyx is an open-source AI platform that connects to your company's docs, apps, and people, enabling enterprise search and AI-assisted knowledge retrieval across 40+ internal data sources. Used by hundreds of thousands of users including Netflix and Ramp, backed by Khosla Ventures and First Round Capital.

poetiq is an AI research and engineering company building reasoning systems and self-improving intelligence infrastructure, founded by former Google DeepMind researchers.

Reer is an early-stage AI startup building the agentic layer for CAD and BIM, with AI agents that run natively inside tools like Rhino and Revit to automate batch edits, parametric design, rendering, and documentation via natural language. Founded in 2024 by Connie Hui (CEO, Harvard GSD) and graduated from Harvard Innovation Labs, the company is in free beta with 9,200+ users targeting the AEC software market.

Superlinked builds information retrieval infrastructure for AI applications, combining multimodal and structured signals to improve enterprise search and matching.

Vellum builds an AI product development platform to design, evaluate, deploy, and monitor mission-critical LLM workflows for enterprise teams.

/dev/agents is building a next-generation operating system purpose-built for AI agents, founded by former Google and Android executives who believe a fundamentally new OS is needed to realize the full potential of agentic AI. The company is led by Hugo Barra, David Singleton, Ficus Kirkpatrick, and Nicholas Jitkoff, who collectively helped shape the Android ecosystem. Valued at $500 million after a $56 million seed round led by Index Ventures and CapitalG, the company is developing foundational infrastructure for how AI agents interact with the digital world.

1001 is a GCC- and London-based deep-tech startup building a localized, sovereign AI operating system for critical physical infrastructure — aviation, ports, shipping, and energy. Its platform builds live operational models of complex systems to predict failures, automate high-stakes decisions, and optimize physical workflows. Founded by ex-Scale AI operator Bilal Abu-Ghazaleh, it embeds forward-deployed engineers with Gulf infrastructure operators.

14.ai is a Y Combinator (W24) company building an AI-native customer-service agency that takes over a company's full support operation across email, phone, chat, WhatsApp, Telegram, and social channels using its own AI systems rather than only selling software. It has since expanded into an AI engine that powers 'autonomous brands,' starting with its own consumer gummy brand gloglo.com. Founded by married duo Michael Fester (ex-Snips, acquired by Sonos) and Mari Schneegans.

Acedly is a real-time AI interview assistant ('interview copilot') that overlays private, low-latency guidance to job candidates during live video interviews across major meeting platforms. Its product suite includes mock interviews, meeting transcription/translation, coding-interview support, and ATS-oriented resume review. It is a consumer-facing tool sold directly to individual job seekers via free, monthly ($69/mo), and lifetime ($988) tiers.

Adaline provides an LLMOps platform for product and engineering teams to iterate, evaluate, deploy, and monitor prompts and AI applications, helping teams manage experimentation, observability, and reliable production rollouts.

AI platform providing reinforcement learning operations (RLOps) tools that enable organizations to customize, evaluate, and serve specialized large language models for production use cases. Founded by key contributors to the Falcon and BLOOM open-source LLM projects.

Adept AI is a machine learning research company building AI systems for human-computer collaboration in enterprise workflows. After raising $415M, the company underwent a reverse acquihire by Amazon in June 2024, with founders joining Amazon while Adept continues operating independently.

Aether AI is a frontier AI research company building causal world models — systems that reason through causal mechanisms rather than statistical correlations, so machines can reason under interventions and act reliably in complex real-world environments. Founded by UC San Diego professor and causal-discovery researcher Biwei Huang, its initial commercial focus is Physical AI and robotics. The company raised a $20M seed round in June 2026 led by MPCi.

Altan is a Barcelona-based startup building an agent-native platform where AI-powered squads autonomously design, build, and operate software from end to end. The platform has attracted over 25,000 builders and businesses, enabling anyone to create websites, web applications, and full software products through conversational prompts.

Anytime AI builds an all-in-one agentic AI platform purpose-built for plaintiff law firms handling complex litigation such as medical malpractice, personal injury, and trucking accident cases. The platform automates legal workflows from intake through settlement, including medical record analysis, demand letter drafting, discovery responses, and trial preparation. Founded in 2023 and headquartered in White Plains, New York.

Neuro-symbolic AI company developing foundational model for conversational agents combining generative AI with rule-based AI for controllable interactions

Autonomous Cyber is an AI-native platform for offensive cybersecurity operations, offering FUZZ-E to support the full cyber kill chain from vulnerability discovery and exploit development to penetration testing and operationalization. The platform augments government and commercial security teams, compressing multi-day tasks into minutes. It was founded by former NSA and U.S. Army personnel and targets both defense and private-sector customers.

BenchFlow is a San Francisco-based frontier environment lab building a unified platform to evaluate AI models and agents using standardized, reproducible benchmarks derived from real-world tasks. It ships products such as SkillsBench, ClawsBench, and the BenchFlow runtime, supporting evaluation of coding agents, web agents, RAG systems, and call-center automation. Founded in 2024 by CEO Xiangyi Li, the company offers custom benchmarks built on reinforcement learning and open community contributions.

Born is an AI character company based in Berlin and New York City (formerly known as Slay). It builds AI companions designed for emotional connection and social play; its flagship app Pengu is a virtual pet that friends co-raise together, with over 15 million registered users. Since 2022 the company has launched five consumer social apps that reached #1 globally on the App Store.

Bounti builds an AI sales agent platform for prospecting, account research, and outreach workflows. The platform automates identifying sales prospects and generating personalized outreach, helping sales, marketing, and customer success teams engage more efficiently while keeping humans in the loop.

Brainfish builds ambient AI support agents and product knowledge systems that help companies resolve customer issues proactively inside their software. The platform focuses on automated answers, product guidance, and lower support burden for support teams.

BuildPrompt is an AI-powered document analysis platform for the construction industry. It enables professionals to extract key information from project documents through natural language prompts, with features for custom AI workflows and model tuning.

ClarityCare AI builds AI agents for utilization management and prior authorization automation in healthcare. The platform processes authorization cases within minutes, applies clinical guidelines with human-level accuracy, and integrates directly into existing utilization management environments. It targets health insurers and healthcare operators to streamline clinical review workflows and reduce UM operational costs.

Common Knowledge is a research lab and product company reimagining the document in the age of AI, building document-native interfaces and workflows for intelligence-rich, collaborative knowledge work.

Contextual AI provides an enterprise platform for building specialized RAG 2.0 agents that enable domain-specific AI applications with advanced retrieval-augmented generation technology.

Continua develops a Social AI agent that integrates into group chats on SMS, iMessage, and Discord to enhance collaboration and productivity. Founded by former Google Distinguished Engineer David Petrou, the AI assistant can automatically set reminders, launch polls, and add calendar invitations during group conversations.

Converge Bio develops generative AI platform trained on molecular data to help pharma and biotech companies discover and develop more effective drugs faster, utilizing Large Language Models specifically trained on biological languages.

Creao AI is an agentic AI platform where AI super agents autonomously execute tasks, write code, call APIs, and deliver reusable automation workflows. Founded in late 2024 and headquartered in Cupertino, California, the platform targets enterprise teams and individuals who want to turn workflows into intelligent, interactive software.

Crescendo is an AI-powered customer experience platform that blends conversational AI with human-in-the-loop support teams to deliver guaranteed CX outcomes. It operates as a managed service, combining AI assistants with outsourced customer operations across retail, financial services, and health & wellness. The company acquired PartnerHero to scale operations to six continents.

Enkrypt AI provides an AI security and governance platform that helps enterprises secure, monitor, and enforce compliance across generative AI and agentic AI systems.

Evolvent AI builds data infrastructure for self-evolving AI agent systems, producing high-quality RL/SFT post-training data and simulated environments for coding, SWE, terminal, and long-horizon agent tasks. Its platform lets a population of agents inside an organization co-evolve over time by accumulating skills and shared memory across trajectories. The team, drawn from top universities and leading foundation-model groups, sells post-training data and agent evaluation infrastructure to frontier model labs.

Fastino builds task-specific language models and small-model infrastructure for production AI workloads, focused on speed, accuracy, and agentic AI use cases.

First Voyage is a New York consumer-AI company behind Momo (Momo Self Care), an AI pet you care for that uses gamified habit-building to improve sleep and self-care alongside personalized self-care chat. Users have created over 2 million tasks on the platform. The company raised a $2.5M seed round in December 2025 to launch Momo on Android and deepen its AI interactions.

Fixie (now Ultravox) builds real-time, speech-native voice AI models and developer tools for building fast, fluent conversational voice agents.

Flint is an AI-powered personalized learning platform for K-12 schools, enabling teachers to deliver individualized instruction at scale through dynamic worksheets, adaptive assessments, and interactive learning activities. Founded in 2023 and backed by Y Combinator, the platform serves 500,000+ users across hundreds of schools globally, including a partnership with Cognita to reach 95,000 students across 110 schools. Flint raised a $15M Series A in November 2025 co-led by Basis Set Ventures and Patron.

Glue is an open-source interface design canvas for AI agents, giving coding agents an infinite canvas to create new interfaces, iterate on existing ones, and turn changes into production-ready code in real time. It is designed as a collaborative workspace where builders and AI agents work together in a single design-to-code workflow.

Go Autonomous is a Copenhagen-based autonomous commerce platform that uses AI and NLP to extract intent from inbound emails, automating quote and order processing for B2B sellers. The platform integrates directly with ERP and CRM systems to enable end-to-end automation of B2B sales operations, reducing manual processing time from hours to seconds.

Guardrails AI builds safety, validation, and monitoring infrastructure for large language model applications. Its open-source and enterprise platform helps developers enforce output checks, reliability constraints, and policy guardrails around AI systems. The platform supports building, governing, and scaling production generative AI across any LLM and deployment environment.

Hercules AI (formerly ZERO Systems) is an enterprise AI company whose platform uses multi-agent AI systems and LLMs to automate complex back-office workflows, turning messy documents and data into real-time, auditable flows for regulated industries. It is especially strong in legal, counting roughly a third of the top 100 U.S. law firms plus Fortune 500 customers such as State Farm and S&P Global. The company raised a $26M Series B in July 2024 led by Streamlined Ventures.

Humoniq is an AI-native BPO platform for the travel and transport industries, combining voice and text AI agents with expert human agents to handle customer support operations including flight changes, rebookings, and cancellations. Targets airlines, airports, online travel agencies, and travel management companies, which collectively outsource $20B of customer support annually. Founded by the team behind Flightfox (YC S12); YC S25.

Ineffable Intelligence is a London-based AI research lab founded in late 2025 by David Silver, former reinforcement learning lead at Google DeepMind and creator of AlphaZero. The company is building a 'superlearner' system that discovers knowledge purely through self-directed experience and reinforcement learning, without relying on human-generated data. Its stated mission is to make first contact with superintelligence.

Instill AI is a persistent AI workspace that transforms complex documents into actionable insights. Users can upload unstructured data such as financial reports, legal contracts, and research papers, ask questions across entire projects, and receive answers traced to specific sources with no coding required.

Kaigo Health builds an AI-powered outpatient care workforce — autonomous voice agents that follow up with, monitor, and advocate for Medicare patients through accessible phone calls designed for older adults, then report actionable insights back to human care teams. The platform automates remote care management (post-discharge follow-ups, ongoing check-ins, gap closure, pre-visit planning) so providers can scale outpatient care without adding headcount. Founded in 2025 and based in San Francisco, it is part of Y Combinator's Fall 2025 (F25) batch.

Lama AI is an AI-native, GenAI loan origination platform that helps community and regional banks and credit unions lend to small businesses, automating the process from application to closing. Its platform is in production at dozens of banks and has processed billions of dollars in loan volume. The company is headquartered in NYC with an R&D center in Tel Aviv.

Go-to-market platform with agentic AI workflows that helps sales and marketing teams automate outbound campaigns, prospecting, and repetitive GTM execution.

Lemni is an AI platform that enables businesses to build fully autonomous agents for customer interactions. Agents can handle email, phone calls, webhooks, and integrations with tools like Shopify, Stripe, and HubSpot.

Liminary is an AI-native knowledge management platform that allows knowledge workers to save and retrieve research from multiple sources including web pages, PDFs, videos, emails, AI chats, and local files. It builds a personal knowledge graph to surface actionable insights for consultants, strategists, and professionals who need accurate, grounded research recall.

Mail0 (also known as Zero) is an open-source, AI-native email client that uses artificial intelligence to prioritize emails, summarize threads, draft replies, and organize inboxes into smart categories. A Y Combinator X25 batch company, it focuses on privacy-first email management that helps knowledge workers save time on email triage.

Mantis Biotech builds infrastructure powering human-in-computer models, combining LLMs with high-fidelity physics simulations to convert rare human behavior data into predictive digital twins. The platform unifies fragmented biological, performance, and operational data into a single interactive system. Initially focused on sports health and human performance, the technology applies broadly to any domain where decisions are critical and data is complex.

AI model routing and interpretability platform using breakthrough Model Mapping technology to unpack LLMs into interpretable architectures. Intelligently directs prompts to best-suited models.

Miso is a personal agentic travel AI that enables high-frequency travelers to curate, book, change, and cancel flights, hotels, rental cars, and experiences through simple text messages. The platform maintains contextual awareness of previous bookings, preferences, and requirements to deliver a personalized travel management experience. Miso searches both cash and award inventory to optimize the cheapest mix for every booking.

Mistral AI is a French AI company building fast, open-source, and secure large language models for developers and enterprises. Known for efficient, high-performance models offered via API and for self-hosting.

AI research company building generative video models and tools for creative professionals, including its licensed Marey video model.

Nexad is the first AI-native advertising system that embeds context-aware, real-time ads directly into AI chat and search applications. Instead of traditional banner placements, it analyzes conversations in real-time to serve personalized suggestions that achieve 2x higher conversion rates than traditional ads.

Okahu provides AI observability tooling to test, monitor, and improve LLM and agent behavior in production systems.

ORO is a human intelligence protocol that enables individuals to contribute private data to AI model training while preserving privacy through zkTLS and TEE cryptography. Users earn token rewards for their data contributions.

Reinforcement fine-tuning platform (YC W25) that helps companies build task-specific AI models that outperform foundation models at a fraction of the cost, applying GRPO-style post-training without managing infrastructure.

Pactum is an autonomous negotiation platform that deploys AI agents to conduct commercial supplier negotiations at scale for large enterprises including Walmart, Maersk, and Wesco. The platform automates procurement, sales, and partner management workflows, handling negotiations end-to-end without human involvement. Founded in 2019 in Estonia, it is one of the earliest production deployments of agentic AI for revenue and spend negotiations.

PaleBlueDot AI develops an AI platform and infrastructure stack for building and deploying AI applications, including agent workflows and model-powered products.

Polsia is an autonomous multi-agent AI platform that plans, codes, markets, and manages businesses end-to-end. It deploys specialized AI agents covering orchestration, social media, email outreach, customer support, ads, finance, business planning, competitor research, and code generation. Customers pay $49/month plus a 20% revenue share on platform-generated economic activity.

PromptQL is an agentic AI data-access and analysis platform built by the team behind Hasura, letting enterprises deploy LLM-powered agents that query, reason over, and act on their business data reliably. It combines a query planning layer with LLM workflows to deliver trustworthy automated data analysis for business operations. The company is San Francisco-based and operates as the AI-native successor product of Hasura.

Ralo is an AI-native online mortgage broker that automates the loan shopping and origination process to offer borrowers rates roughly half a point below the national average and faster closings. Founded in 2025 and based in New York City, it is a Y Combinator-backed company currently licensed in California, Colorado, and Texas.

South Korean AI chip company developing neural processing units (NPUs) for AI inference workloads. Offers AI accelerators, chiplets, interconnects, and software stacks for large-scale deployments.

Reworkd builds AI-powered web data extraction agents that automate the entire scraping pipeline at scale. Note: Company announced sunsetting as of February 2025.

Riza provides isolated code execution infrastructure for AI-generated code and agent workflows.

RunPod is a GPU cloud infrastructure provider that enables developers, researchers, and AI companies to deploy and scale GPU workloads on demand. The platform offers serverless inference, on-demand GPU pods, and instant multi-node clusters, serving over 500,000 developers.

Safe Superintelligence is an AI research company founded by Ilya Sutskever to pursue a singular mission of building safe superintelligence.

Shofo is building the world's largest indexed video library — described as 'Common Crawl for Videos' — by continuously crawling and labeling public video content from social platforms like TikTok. It sells custom, multi-stage annotated datasets (object detection, activity recognition, semantic segmentation, reasoning annotations) to AI labs for pre-training and fine-tuning. Part of Y Combinator's W2026 batch.

Sonatic is a promptless agentic actions platform that functions as a desktop assistant observing your screen, learning how you work, and autonomously completing routine tasks across workplace applications. Built as a new layer of personal computing, Sonatic turns intent into action without requiring explicit commands or prompts.

Strawberry is a Stockholm-based startup building the self-driving browser — an AI-powered web browser with built-in autonomous agents that can surf, click, and perform real multi-step tasks on behalf of users, even on login-protected sites. Designed for non-technical professionals like salespeople, recruiters, and analysts, Strawberry eliminates the need for advanced prompting or coding to automate complex web workflows. Backed by General Catalyst and EQT Ventures, the company has moved from closed beta to open availability.

Universal memory API and context engineering platform for AI applications. Provides long-term and short-term memory infrastructure for AI agents, powering enterprise APIs, developer plugins, and a personal knowledge app.

Symbolica is a pioneering AI research lab founded in 2022 that leverages category theory to deliver breakthroughs in modern artificial intelligence research by building structured reasoning in machines with formal logic and semantics at the architectural level.

TensorZero is an open-source stack for industrial-grade LLM applications with tooling for evaluation, observability, and production optimization.

Tilde Research is an AI infrastructure company building efficient mixture-of-experts (MoE) training and inference systems. It develops novel architectures and hardware-aware algorithms to reduce the cost and improve the performance of large language model training.

Topify helps brands measure and improve visibility in AI-native search experiences such as ChatGPT, Gemini, and Perplexity with analytics and optimization tooling.

TraceRoot AI is an open-source self-healing layer for AI agents that automatically identifies and resolves production bugs. It captures traces, logs, and metrics, then uses AI agents that analyze source code and GitHub history to perform root cause analysis and auto-fix issues. Part of Y Combinator's S25 batch.

Traini is a Palo Alto-based startup building generative-AI apps and a Cognitive Smart Collar for pets. Its models interpret a dog's emotions and intent from barks, movement, expressions, and physiological signals to enable real-time human-dog communication, health monitoring, and AI pet training. Its emotion-translation accuracy reaches up to 94% across nearly 120 breeds.

Veris AI is a simulation-first training and certification platform for enterprise AI agents. Their platform provides high-fidelity sandbox environments where AI agents can be stress-tested, trained via reinforcement learning, and graded against accuracy, cost, and latency SLAs before deployment to production.

Vesence builds agentic AI agents embedded directly into Microsoft Office (Word, Outlook, Excel, PowerPoint) for law firms and professional services. The platform provides document quality assurance, compliance checking, and drafting acceleration by running hundreds of AI agents inside familiar Office tools. Founded by Henrik Hansson and Ludvig Swanstrom (formerly of YC-backed Depict), Vesence is backed by Emergence, Creandum, Y Combinator, and 20VC.

Vulcan Technologies uses AI to map every law, regulation, and court case in America, enabling government agencies and businesses to replace massive consulting projects with instant automated analyses.

Wordware is a web-hosted IDE for building AI agents, enabling non-technical domain experts and engineers to collaboratively create, iterate, and deploy task-specific AI applications 20x faster. Backed by Spark Capital, Felicis, and Y Combinator, the platform offers prebuilt tools, API deployment, and tracing.

Wrtn Technologies builds consumer and business AI services that make generative AI tools accessible in everyday work, content creation, marketing, and agent-style application experiences across Korea and Japan.