Vellum, the New York-based upstart with enterprise muscle and MIT-grade IQ, just closed a $20M Series A led by Leaders Fund, with backing from Socii Capital and returning believers Y Combinator, Rebel Fund, Pioneer Fund, and Eastlink Capital. That’s $25M total fueling a company that isn’t playing the “scale fast, break stuff” game. They’re writing test-driven workflows for AI the way software should have been built, secure, observable, and actually usable by teams that don’t speak in tensors.
Shoutout to co-founders Akash Sharma (CEO), Sidd Seethepalli (CTO), and Noa Flaherty (COO), the kind of trio that could turn a whiteboard sketch into a revenue engine. Which they did. $3.5M booked in 2024. Customers like Drata, Swisscom, Redfin, and Headspace aren’t showing up for hype. They’re betting on a platform that gives you continuous integration / continuous deployment for large language models, test coverage for prompts, and governance you don’t have to explain twice to compliance.
Vellum came out of a real-world itch. Back at Dover (YC S19), the team hacked together tools for debugging large language model workflows, tools that should’ve existed but didn’t. So they built them. Now, in Q1 2025, their platform’s GA, enterprise-grade, and making artificial intelligence product launches 10× faster with fewer faceplants. Turns out, if you treat AI workflows like actual software, things don’t fall apart in production.
Here’s what makes it sing: visual workflow builder with real branching logic, SDK with versioned releases, native evals that break down latency and cost like a surgeon, and rollbacks that don’t make you pray. And yeah, they plug into OpenAI, Anthropic, Meta, retrieval-augmented generation pipelines, and whatever backend you’re wrangling data from. All wrapped in containerized, cloud-agnostic infrastructure and built with role-based access controls and encryption that doesn’t mess around.
Series A money’s going straight into R&D velocity, EMEA and APAC expansion, and scaling the go-to-market machine. By Q4 2025, they’re planting a London HQ. By 2026, they want 500 enterprises in the fold and a headcount tripled to 150. And they’re not just chasing logos, they’re digging deep into financial services, healthcare, and every compliance-obsessed vertical still pretending spreadsheets are workflows.
Vellum isn’t painting inside the lines. They’re redrawing the canvas. Congrats to Akash Sharma, Sidd Seethepalli, and Noa Flaherty for building what enterprise AI actually needed: structure without friction, scale without chaos.
