Engineering, Product and Design Startups funded by Y Combinator (YC) in the San Francisco Bay Area that are currently hiring 2026

Scale accelerates the development of AI within organizations of any size to deliver critical business insights and operational efficiency. Its data-centric infrastructure platform leverages RLHF (Reinforced Learning with Human Feedback) to help organizations build the strongest AI models that supercharge their business, with customers across industries including Meta, Microsoft, U.S. Army, DoD’s Defense Innovation Unit, Open AI, General Motors, Toyota Research Institute, Brex, Instacart and Flexport.

Runtime (runtm.com) gives engineering teams control and visibility to enable anyone in their org to use coding agents securely. Eng leaders rolling out coding agents have no way to see what's happening, set boundaries, or safely let non-engineers use them. We provide sandboxed environments, session-level observability, and configurable guardrails so anyone on your team can ship with coding agents without breaking prod. We're live with fast-growing YC scaleups empowering their Engineers, PMs, designers, marketers, and ops to contribute to their codebase or build prod-grade zero-to-one apps.

Replicas is a background coding agent that allows you to delegate tasks to Claude Code or Codex from Slack, Linear, GitHub, and more. Each task works inside of a sandboxed virtual machine, allowing teams to heavily parallelize work and burn through tickets. Replicas agents output higher quality PRs via feedback loops - each task automatically is notified and fixes CI failures and reads all code reviews. With Environments, engineering teams can configure agents to run their entire codebase locally, verify its work, and provide preview URLs.

There are 2 problems: The first, CAD software is slow and the second, designing with it is a tedious process with many steps. Aurorin solves both of these problems by being a CAD software built from the ground up on a modern tech stack with AI at its heart. Building the CAD kernel lets us fully leverage AI while modernizing execution speed. Almost all existing CAD software (Solidworks, NX, Creo, Catia, Fusion 360, Onshape etc) use the same couple of underlying CAD kernels built in the 80's. It's common for these programs to take 4 hours to open a file. Aurorin uses a custom parametric and B-Rep driven CAD Kernel built to take full advantage of modern CPUs and GPUs and be AI native.

Sparkles is like Lovable for existing projects, so even Sharon from Marketing can change the button styling on the front page without going back and forth with the dev team. Someone technical connects their GitHub to Sparkles, sets up the repository and environment variables, and then anyone on the team with a @yourcompany.com email can log in on Sparkles and create pull requests without needing to know what Git is.

AtlasGrid automatically builds a visual graph of your mobile app UI and displays analytics directly on your screens. Additionally, our fleet of AI-powered agents will continuously analyze thousands of mobile apps, gathering design insights and identifying growth opportunities + actionable product improvements.

Antigen is an AI adversary for the enterprise. It maps, exploits, and secures your attack surface through nightly automated offensive testing and monthly human red team assessments. Every finding is delivered with traces, reproduction steps, and remediation directly in GitHub, Linear, or the Antigen Platform. We work with companies in financial services, healthcare, retail, and government where security is trust. We're backed by Lightspeed, DST Global, and Susa Ventures.

We work with frontier AI labs to help train their agents to become AI research scientists

Floot lets non-coders & entrepreneurs build real web apps that actually work. We’re taking a new approach by building everything from the ground up for AI. ✨ Try building an app at floot.com Existing AI app builders fall short of enabling non-coders to build real, serious apps. They often: - Break down when building anything more than a prototype or toy app - Have non-coders stuck with errors they don’t understand - Require non-coders to handle multiple services designed for developers (e.g. for backend, database, hosting, etc) The fundamental problem: these AI app builders just cobble together existing developer tools and slap AI on top, but these tools were designed for humans and not for AI. The result? Non-coders get excited, start building, hit a wall, and then abandon their projects. Floot is purpose-built for AI. Our secret sauce: 1) Instead of reusing a coding framework designed for humans, we created a specialized framework that dictates how AI should write code and how the code runs on our infrastructure. This allows us to enforce best practices, detect errors earlier, and give the best feedback to the AI. As a result, Floot can write better code and fix issues more effectively. 2) Floot has the entire tech stack built-in (backend, database, hosting, etc) so building and managing an app feels seamless, especially for non-coders. And because we control every layer of the stack, we can optimize each layer for AI-first coding. Because of this, non-coders can get more done with less time and prompting. They can finally build serious apps that work, instead of getting stuck in frustrating error loops and then abandoning their projects. We also built a highly scalable hosting system directly on AWS so Floot apps can handle real traffic and business demands. This means non-coders can build, deploy, and scale their apps on a single platform seamlessly. Plus, Floot is really fun to use! Try building an app at floot.com

TraceRoot is an open-source observability and self-healing layer for AI agents. It captures traces, debug with AI that sees your source code and GitHub history.

Embedder is building an AI coding agent that can write, test, and debug firmware on real hardware.

BootLoop is building the first AI agent to write and test firmware on real hardware. Existing AI coding assistants fail completely with embedded systems, but our agent ingests datasheets, schematics, and other design files to write firmware in your style and automatically test it on your hardware.

Wafer builds AI agents that work as autonomous performance engineers, optimizing GPU kernels for AI inference. Our product is serverless and dedicated inference for the world’s fastest open source LLMs, achieved by Wafer's autonomous performance engineers.

Autosana is the end-to-end testing layer for your coding agents, across iOS, Android, and Web apps. Step away from your laptop. Come back to full-stack features built and tested end-to-end, production ready. Autosana creates, updates, and runs tests automatically from your code diffs. Locally, it loops with your coding agent until every test passes. In PRs, it loops with cloud agents and shows video proof of your new feature or bug fix working end-to-end. No setup, no maintenance, just tests evolving with your codebase. Ship with confidence.

We are a product lab solving context in AI.

Idler builds reinforcement learning environments that teach AI models to code at expert human levels. We create training environments based on real-world coding scenarios that prepare models for the complex challenges they'll face in production.

Manage a fleet of agents locally or in the cloud. 100% open source and free to use. Built to help you do your life's work.

Theorem is training models that make program verification 10,000 times faster. Using verification as a feedback loop, developers have found zero-days in GPU accelerated code and cryptography implementations, and sped up code migration in legacy systems. If you have complicated code that needs to be correct and secure, sign up for our beta!

What if every experiment you wanted to run could be modeled using AI? Godela is an AI-powered physics engine that gives engineers faster, cheaper replacement to simulations and physical prototypes. No set up, no wait. Just ask—and get simulation quality result.

StarSling is building an AI-native DevOps platform for developers that ship, starting with CI. StarSling Runners are fast GitHub Actions runners with AI agents that continuously improve your build speeds by more than 80%.

Attune makes blazingly fast build tools that require zero migration effort. As AI enables developers to write code faster than ever, slow builds are increasingly becoming the bottleneck in development velocity. Developers regularly lose hours of productivity every day waiting for their builds. Imagine being able to write code in 1 minute, but needing to wait 30 minutes to compile, deploy, and test it. Existing solutions either require painful migrations to new tools (e.g. Bazel) or rely on just using bigger servers to run your builds. Instead, we do native, first-class integrations into compilers and build systems. With Attune, you don't need to integrate a new tool. Just make a one-line change from `cargo build` to `hurry cargo build` for 20x faster builds. Eliza (ex-Google) and Xin worked together for 3 years at FOSSA, where Eliza was Head of Engineering and Xin was Head of Product. Eliza brings a decade of programming languages and build systems expertise, while Xin brings a decade of experience building developer tools products, most recently as VP of Product at Teleport (S15). Having experienced the pain of slow builds firsthand across hundreds of production codebases, we're building the solution we've always wanted.

Give every agent a cloud desktop. Built for Claude Code, Codex, OpenClaw, and any computer-use agents.

Partcl modernizes chip design automation with physics-informed models powered by GPU acceleration. Our tools run up to 700× faster than legacy solutions, cutting weeks off development and unlocking AI-driven optimization.

HUD (YC W25) is developing agentic evals and RL environments for Computer Use Agents (CUAs) that browse the web for frontier AI labs. Our CUA Evals framework is the first comprehensive evaluation tool for CUAs. People don't actually know if AI agents are working reliably. To make AI agents work in the real world, we need detailed evals for a huge range of tasks. We're backed by Y Combinator, and work closely with frontier AI labs to provide agent evaluation and training infrastructure at scale.

Weave uses AI to understand engineering.

Wildcard is the AEO and GEO platform designed specifically for e-commerce and retail brands. We help merchants improve visibility across AI search and AI shopping platforms like ChatGPT, Google AI Overviews, Gemini, and Amazon Rufus. Wildcard tracks how brands, categories, collections, and SKUs appear across AI search results, identifies visibility gaps, and turns those insights into actions. We use AI agents to enrich product data, generate SEO, AEO, and GEO content, create collection and comparison pages, build FAQs, and improve off-site discoverability across Reddit, YouTube, blogs, and other sources. As AI shopping evolves, Wildcard is building the infrastructure brands need to win in agentic commerce, including support for emerging standards like the Agentic Commerce Protocol (ACP) and Universal Commerce Protocol (UCP).

Onlook is the fastest way for anyone to build, deploy, and iterate on websites and web apps. We built an open-source, visual editor for code that merges design, development, and AI. Since launching, Onlook hit #1 on Hacker News, was the #1 trending repo on GitHub—above DeepSeek—and has earned over 8,500 GitHub stars. Founded by Daniel Farrell, a designer of 10+ years and former Head of Growth, and Kiet Ho, an ex-Amazon and ex-ServiceNow engineer, Onlook is on a mission to finally solve the gap between design and development.

a0.dev is a platform that helps mobile developers build high quality apps in days instead of weeks by generating custom react native projects. We've launched top ranking apps that have been used by hundreds of thousands, and now we're creating the tools we wish we had to build faster.

Adam is building AI CAD. It transforms text into parametric designs so engineering teams can ship 10x faster. Our mission is to eliminate the gap between engineering and reality so more great ideas come into existence.

Dalus is an AI-powered systems engineering software to model/validate complex systems for aerospace/defense/robotics and other mission-critical industries. With Dalus, engineers can seamlessly trace requirements, perform analyses, and update information in real-time. This ensures that every team member stays aligned and informed, reducing inefficiencies and enabling faster iterations. With AI at our core, users can analyze hundreds of pages of existing engineering documentation and requirements to generate or retrieve information about the system model they are building in Dalus. We started working together on Dalus in spring 2024 after Eliot quit his job at Aerojet Rocketdyne (L3Harris) where he worked as a test/systems engineer. He experienced first-hand how silos and poor communication reduced the productivity and efficiencies in the teams he worked at. We love working with companies tackling complex engineering systems, whether they be startups just getting their feet wet or large organizations like Israel Aerospace Industries (Our design partner and biggest aerospace/defense company from Israel) Read why we started Dalus: https://dalus.io/articles/why-dalus

Confident AI allows companies of all sizes to benchmark, safeguard, and improve LLM applications, with best-in-class metrics and guardrails powered by DeepEval. Built by the creators of DeepEval (12.6k stars, >3m monthly downloads), Confident AI is able to offer battle-tested, open-source evaluation algorithms while providing the infrastructure needed for teams to stay confident their LLM systems.

Docs are integral to a great customer experience, but managing them is a challenging, multi-team effort. The most successful companies recognize that fantastic docs drive adoption and bad ones cause customers to churn. Promptless is used by fast-growing startups and Fortune 500 enterprises alike to automatically update docs based on feature releases, support tickets, internal Slack conversations, and more. Promptless is designed to be flexible—it can integrate with any docs platform, be triggered automatically from many workflows, and can pull in context from any internal knowledge source. With Promptless, documentation becomes a strategic advantage for growth, not a bottleneck or a liability.

Creative intelligence

The world's best AI web app builder, implementing a radically unique architecture that lets us completely outperform the competition in complexity, reliability, scalability, and more. We are 7x cheaper and are positioned to take down Lovable, Replit, Base44, Bolt, Emergent, and more.

Orchestrate an army of coding agents from one dashboard. Ship entire sprints in parallel.

Tired of switching tabs to cursor every time you're building your iOS app in Xcode? Worry not! Alex is a sidebar app that works directly with Xcode, and has the same Cmd+L, Cmd+K, and Apply Code commands as Cursor. And, it's super fast. Check it out: https://alexcodes.app

We're making consumer agent products completely free, using the best open source models, and funded by ads. Our first product is a coding agent CLI called Freebuff. Install it with "npm i -g freebuff". We're also working on a free Lovable competitor.

Cekura helps Conversational AI companies with automated QA. From pre-production simulation and evaluations to production call monitoring, to assisting customers in setting up the CI/CD pipeline, we help in QA across the agent lifecycle. We are working with 75+ customers across healthcare, BFSI, logistics, recruitment, and retail. Some of the largest CCAS players and enterprises use Cekura to ship AI voice and chat agents reliably

HumanLayer is an IDE and collaborative cloud platform that helps software engineers solve hard problems in complex codebases. Ship 2-3x faster without AI slopping up your codebase.

Unsloth helps builders create custom models better & faster using RL and custom kernels. We're developing the all in one solution to help you create highly-accurate custom models 30x faster with 90% less memory use. With over 10 million monthly model downloads and 40K GitHub stars, our mission is to make training and open-source the best it can be!

Coval is a simulation & evaluation platform for autonomous AI agents, helping engineers launch dependable assistants across chat, voice, and other modalities. We simulate thousands of scenarios engineers don't have to manually test their agents. Our CI/CD evaluations automatically simulate and detect regressions.

Run parallel agents in isolated workspaces. See at a glance what they’re working on, then review and merge their changes.

DeepSim is building an AI physics simulator. We are currently developing the only thermal simulator to meet AI chip design needs and are validating our tool with Intel. We are a team of three electrical engineering PhDs from Stanford with backgrounds in semiconductor fabrication and design.

Helium is building self-improving software — AI that continuously experiments on key moments in the user funnel to boost revenue. Our first customer, a social app with 20M MAU, uses Helium to test new in-app paywalls to get more users to subscribe. Zach previously made $100M+ in growth experiments at Uber, Shishir led the first LLM features for Alexa, and Anish worked at Meta building ML to automatically experiment on ads, increasing Instagram ads conversion by 10%.

Hestus is AI-powered CAD software that speeds up hardware development. While engineers are typically quick at coming up with initial design concepts, they are bogged down getting from idea to the final design due to countless tedious and repetitive tasks. We are building CAD software that automates these mundane tasks and speeds up the design execution process, while letting engineers spend more time on the creative part of the design process.

Building out new LNG plants, HVAC systems, or semiconductor processes rely on hundreds of iterations of feasibility testing (through simulation or real-world lab tests) where different designs (combinations of equipment) and parameters are validated. The parameters are often locked in PDFs (datasheets, wiring diagrams, PFDs/P&IDs) which take thousands of hours to convert into CSVs / JSONs, and the simulators don't have an easy API to work with. Outerport bridges the gap by finding documents from PLMs, extracting structured data from drawings, building a knowledge graph over them, and building autonomous AI agents that can fully automate this R&D process by running simulations and performing design checks.

Cerulion is an open source operating system for robots built by 2 MIT robotics PhDs. Companies like Amazon Robotics and Boston Dynamics AI Institute are using us to develop faster and ship more reliable robots than with ROS. With up to 1000x improvements in communications performance over state of the art, Cerulion is enabling the next generation of embodied AI.

Domu automates debt-collection calls, texts, and emails for financial services and healthcare. We work with some of the largest financial institutions in the world and win by doing what nobody else is willing to do. We are hiring :)

Saphira is the easiest way for hardware products to get safety certified and reach market faster, such as industrial robots and heavy machinery.

Entangl uses AI to detect and automatically resolve issues in data center engineering and operations, ensuring efficiency and uptime.