Agents & Orchestration Founders

Aitken made two iPad apps that won Apple's iPad App of the Year before co-founding Trigger.dev, an open-source platform that runs long background jobs and AI agents written in TypeScript.

Allam co-founded and ran engineering at Code School, an early learn-to-code site, and now leads engineering at Trigger.dev. His team rewrote that open-source background jobs platform twice before it found steady growth.

Arendtsz watched companies drown in disconnected SaaS apps after years inside NetSuite, then rebuilt Celigo so business teams could wire those apps together without waiting on a specialist for every flow.

Bearden has spent two decades turning open-source software shifts into enterprise companies, and at Sema4.ai he is trying to move large companies out of AI pilot projects and into agents that do production work.

Bedi spent more than 15 years building data and machine learning systems at companies like Airbnb and Facebook, then built Agno, an open-source Python framework for AI agents with more than 40,000 GitHub stars.

Behal left Amazon after deciding big tech moved too slowly, then joined Max Brodeur-Urbas after the two presented side by side at an AI event in Vancouver, Canada. As Gumloop's CTO he builds the software that lets nontechnical teams create their own AI agents and automations.

Bhagwat is trying to make TypeScript the default place developers ship agents, with the same docs seriousness the Gatsby team brought to the web.

Bharadwaj is building Dodge AI so agents do most of the upkeep on systems like SAP, work that companies now staff with large consulting teams.

Brodeur-Urbas started Gumloop as a side project in a Vancouver bedroom and turned it into an AI agent builder that employees at Shopify, Ramp, and Instacart use without engineers.

Bryk is building Exa to be the search engine AI agents use, ranking the web by meaning so an agent gets the right pages from one plain-language query.

Burazin has worked on cloud development tools since co-founding Codeanywhere in 2009, and Daytona turns that experience into fast, stateful sandboxes for AI agents.

Chalef built Zep after open-sourcing a tool to summarize chatbot conversations, and its Graphiti framework for temporal knowledge graphs now has more than 30,000 GitHub stars.

Chase built LangChain so building software with language models feels as ordinary as building with any other library.

Cherkasov wants agents to inherit a company knowledge graph so ops teams can route work between people and AI without rebuilding context every time.

Ciarla leads growth at Firecrawl, the open-source API that turns web pages into clean data for AI apps, and he writes out the reasoning behind the company’s biggest turns, from closing Mendable to paying Wikipedia for access.

Ding wants B2B teams to ship customer integrations through one unified API instead of burning engineers on each vendor connector.

Farrelly taught himself to code after starting in civil engineering, rose to CTO of Buffer and then co-founded Inngest. Its software lets developers write reliable background jobs and AI agent workflows as ordinary code.

Fateev built Temporal so long business processes in software survive crashes and retries the way a careful person picks up where they left off.

Foo spent four years on an email app few would pay for, then made Paragon customers pay before any code existed, and Paragon now runs the integrations behind AI products such as You.com and Copy.ai.

Since 2011 Foster has built Zapier on the belief that people who do not write code should be able to connect their apps and automate their own work, and he now runs it as a company where every new hire is expected to work fluently with AI.

Freund taught himself to program as a boy in Berlin and co-founded Camunda as a small process consultancy in 2008. He turned it into a software company whose workflow engine banks, insurers and manufacturers use to run their business processes, and now their AI agents.

Ganatra watched nearly half a startup's engineers get stuck on integrations, and built Composio so AI agents can connect to more than 1,500 apps through one platform.

Garcia is building the low-level systems that let Kernel start cloud browsers for AI agents in milliseconds and keep them parked cheaply for days.

Guldener grew a Zurich fleet software company to more than 1,000 customers, then started Nango after seeing how hard product integrations were to build, and it now runs integrations for Replit, Mercor, and Ramp.

Holdstock-Brown built Inngest so product engineers can ship reliable AI workflows as ordinary code, without babysitting queues.

Hubert argues companies should run AI agents on their own private data and tools, under their own access rules.

Jue is building Kernel so AI agents can open a real browser in milliseconds and work on the large share of the web that has no API.

Jukić came up with Codeanywhere, one of the first cloud code editors, because he did not want to carry his laptop everywhere as a freelance developer. As Daytona's CTO he is turning 15 years of cloud development work into sandboxes where AI agents run code.

Klein is building Browserbase so AI agents can use websites the way people do, on cloud browsers that developers do not have to run themselves.

Koneru had already taken companies through NASDAQ listings and a mobile-platform sale when he founded Kore.ai in 2014 to put conversational AI in front of enterprise customers and employees before the category had a name.

Kumar is trying to give a two-person business the operating leverage of a much larger team, from first site to paying customers.

Liu built LlamaIndex so language models can answer questions from a company's own private files, not only the public web.

Lowin builds orchestration for data and AI workflows that fail in production when someone's cron job silently stops.

Markovic left more than a decade of data engineering to study cognitive science and clinical psychology, and his open-source memory engine applies that research to how AI agents store and recall what they learn.

An Elevation AI partner who looks for enterprise agents with paying customers, as he did when he led Dextr's first institutional round.

Mizrahi is building Linkup to give AI products accurate, cheap web search, with an eye on paying the publishers whose content agents read.

Mlejnsky and his high school friend Tomas turned a developer tools side project into E2B, which runs isolated cloud machines for AI agents and reports more than 2 billion sandboxes started.

Moura built CrewAI so AI work can run as teams of agents with clear roles, not one giant agent acting alone.

Müller built the first version of Browser Use in days with his co-founder at ETH Zurich, and the open-source project became one of the most popular tools for letting AI agents use websites.

Mulyar is rebuilding Nomic around the drawings, specifications, and submittals that engineers still check by hand, after the open-source AI tools that made the company known were absorbed by the large model makers.

Oberhauser built n8n so anyone can wire tools and AI steps into open workflows they host and change themselves.

Packer co-created MemGPT as a Berkeley Ph.D. student, an early system for giving AI agents long-term memory, and now runs Letta, which turns that research into open-source tools for agents that learn.

Partee came to AI through supercomputing, building tools that linked machine learning to climate simulations before leading applied AI at Redis. At Arcade.dev he works on a problem most agent builders hit quickly: letting an agent act for a real user with real permissions.

Pawar is trying to make warehouse robots and human crews share one work brain, so mixed floors stop running as disconnected fleets.

Peffer bet his company on Firecrawl because every AI team needs clean web data and kept rebuilding the same broken scrapers.

Pelaseyed built Superagent so teams can own an open agent stack end to end.

Pithadia is building Valyu so AI agents in finance and science can search filings, papers, and trial data as easily as the open web, with sources attached.

Polu sold his first startup to Stripe, spent five years there, then did research at OpenAI on mathematical reasoning. He left in 2022 convinced that powerful models were barely deployed, and co-founded Dust to put AI agents to work inside companies.

Raman is building Anchor so AI agents can finish the last step of enterprise work on old portals and logged-in websites that were never built for software to use.

His early backing, with Greenoaks' Neil Mehta, valued Instinct at $50 million, months before its $10 billion round.

Ruecker is the engineer of Camunda's founding pair. He has worked on open-source workflow engines for more than 15 years and written the books many developers use to learn process automation, from Real-Life BPMN to Practical Process Automation.

Rusic built deepset so production NLP and retrieval rest on an open framework teams can own.

Salazar built the first authentication API for developers at Stormpath, sold it to Okta, and now runs Arcade.dev, which lets AI agents act in business systems under each user's own permissions.

Shah turned one of the weekly side projects he built after moving to the US into Supermemory, a memory engine for AI agents that now reports more than 100,000 organizations.

Singh is building Skyvern so companies can automate form filling and portal work across hundreds of websites that were never built for software.

Singh built Mem0 so AI agents get durable memory that lasts across sessions.

Stewart turned a teenage hobby of hosting game servers into Northflank, a platform that now runs microVM sandboxes for AI agents alongside apps, databases, and GPUs.

Swerdlow went from coding in elementary school to a consulting job at Apple to Freestyle, which gives coding agents full Linux virtual machines they can fork with their memory intact.

Thambapillai is building the control layer that decides what company AI assistants may read and change, drawing on seven years of putting sensitive data to work at Palantir.

Tian is trying to put agents inside the messaging apps people already open all day, so products stop dying in the app store.

Vaidya met his co-founder at a physics olympiad camp in India and later built distributed systems at Rubrik. As CTO of Composio he runs the layer that lets AI agents sign in to tools such as GitHub, Slack and Salesforce, now handling about 10 million tool calls a day.

Vassilev built Relevance AI so companies can run whole workforces of AI agents on ordinary business processes.

Waldron wants large companies to connect their business software and AI agents through one integration platform, so an automation can span many systems without custom code for each.

Wall led the Google File System team, co-founded Google Wallet, and sold a fintech startup to Stripe before starting Runloop to give AI coding agents their own work environments.

Wooders co-created MemGPT, one of the first open-source projects to give AI agents a memory they manage themselves, while she was a PhD student at UC Berkeley. As Letta's CTO she is building agents that keep learning over time, with memory that works across models.

Yadav led the AI platform team at Tesla Autopilot and built the open-source EvalAI benchmarking platform as a graduate student. He co-founded Mem0 with a college friend to give AI agents a long-term memory, and as CTO he designs how it stores and retrieves what agents learn.

Yudovskiy took ElectroNeek through Y Combinator with a bet that IT service providers, not only Fortune 500 automation teams, should be able to deliver RPA without per-bot license walls.

Zhang is building reinforcement-learning infrastructure so production agents can learn from outcomes instead of repeating the same mistakes forever.

Luyu Zhang is founder and chief executive of Dify, an open-source platform that helps developers build AI applications through a low-code interface. He grew up in Anhui province, China, dropped out of middle school to write code, and moved to the United States in 2025. He works from Menlo Park, California, according to Forbes.

Žunič, a physics graduate from Slovenia, built the first version of Browser Use in four days with Magnus Müller and saw it go viral as one of the tools behind Manus. As CTO he is building what he calls a foundation layer for agents that use websites the way people do.