A computational physicist who spent a decade in data infrastructure decided the least glamorous part of building software - APIs and SDKs - deserved a real product. Speakeasy is the result.
Ask most founders what excites them and you will hear about markets, moonshots, or the next platform shift. Ask Sagar Batchu and the conversation drifts toward webhook management, SDK ergonomics, and the quiet machinery that lets one program talk to another. That instinct - to care about the parts of software everyone else treats as chore work - is the whole reason Speakeasy exists.
Batchu is the co-founder and CEO of Speakeasy, a San Francisco company that takes a single API specification and turns it into production-grade software development kits across nine or more programming languages, along with Terraform providers and, more recently, MCP servers for AI agents. It is unglamorous work by design. It is also work that nearly every engineering team needs and almost none enjoys doing by hand.
Batchu did not follow the standard founder script. He studied physics and computational physics at Harvey Mudd College, graduating around 2015. Physics, not computer science, is the training underneath his engineering career - and it shows in the way he talks about systems: constraints, interfaces, and what can be safely exposed.
His first job out of school was at Enlighted, an Internet-of-Things proptech startup where a fellow Claremont alumnus, Joe Costello, was CEO. When Siemens acquired Enlighted, Batchu did something a little contrarian. Instead of settling into a larger, more stable organization, he moved to LiveRamp's brand-new London office to build an engineering team from nothing.
That London chapter is the part of his resume that best predicts what came next. As Director of Engineering, he grew the dev center from zero to more than fifty engineers working on data infrastructure and privacy technology. Scaling a team across an ocean, on hard problems, taught him that developer experience is not a nice-to-have. It is the product.
Every developer has a bad-SDK story. A poorly generated client library. A batch of integration code written by hand and quietly rotting. An API that technically works but is miserable to consume. These complaints are universal, and precisely because they are universal, they get treated as background noise rather than an opportunity.
Batchu saw it differently. In 2022 he teamed up with Simon Yu - now Speakeasy's Chief Product Officer - to build a developer-first platform around a simple premise: if you write a clean API specification, the rest of the toolchain should generate itself. SDKs, documentation, Terraform providers, tests. The heroics of hand-writing client code should not be required.
By late 2024, the bet was paying off. Speakeasy reported nearly 3,000 users and more than 7,250 SDKs generated on the platform. The customer list read like a checklist of teams with high standards: Vercel, Mistral AI, and Verizon among them. When demanding engineering organizations not only adopt a tool but champion it internally, investors notice.
In 2024, Speakeasy closed a $15 million Series A led by FPV Ventures, with participation from GV (formerly Google Ventures) and Quiet Capital, plus angels including Prisma CEO Søren Bramer Schmidt and Cribl co-founder Clint Sharp. It followed a $7.6 million seed led by GV, bringing the total to roughly $29 million.
The timing turned out to be fortunate. As large language models moved from answering questions to taking actions, a new requirement emerged: agents need well-designed APIs to actually do anything. An LLM that can call tools is only as capable as the interfaces it is allowed to call. Batchu had spent years building exactly that layer.
He describes the near future in plain terms - agentic systems executing complex workflows through API calls, and companies needing their APIs to be readable by machines and models alike. Speakeasy's expansion into generating MCP servers, the connective standard for AI tools, is a direct answer to that shift.
There is a lesson in Batchu's path worth stealing. The most interesting problem is rarely the one that gets talked about at parties. Software's leverage often hides in the 90% that is plumbing - the SDKs, the specs, the webhooks, the tests. Founders who are willing to fall in love with that layer, rather than route around it, sometimes end up owning the ground everyone else has to stand on.
Batchu tends to say Speakeasy is "only getting started." Coming from someone who chose to build in the least fashionable corner of developer tooling - and got Vercel, Mistral, and GV to agree it mattered - that is less a slogan than a pattern he has already run once. He picked a boring problem on purpose, and boring, it turns out, scales.