Sri Raghu Malireddi is hiring a senior or staff SDK engineer for Moss, the YC-backed semantic search startup built around a narrow but consequential claim: conversational AI breaks when retrieval is too slow.
The YC job listing, visible on July 10, describes Moss as a "real-time semantic search layer for conversational AI" and offers $60,000 to $300,000 in salary, plus 0.10% to 0.50% equity. The role is full-time, tagged as engineering and embedded systems, open in San Francisco or remote, requires 6+ years of experience, and says Moss will sponsor visas.
This is not a general AI application job. Moss wants someone to own SDK architecture across JavaScript/TypeScript, Python, Swift, Android, Elixir, C, and the Rust core they bind to. The listing asks for systems engineering experience across memory management, concurrency, FFI, package distribution, build systems, cross-platform development, WebAssembly, mobile runtimes, edge environments, browser internals, search, databases, AI infrastructure, compilers, and developer tooling.
That breadth says Moss is trying to win distribution at the layer where developers first touch infrastructure. As the listing puts it, the SDKs are "how every one of those developers meets Moss," and the work includes shrinking bundle size, lowering cold-start and query latency, tightening cross-platform parity, improving developer ergonomics, and making integration effortless across browsers, servers, mobile devices, edge runtimes, and embedded environments.
What Moss says it does
Moss describes itself as a real-time semantic search layer for conversational AI, delivering sub-10 ms retrieval directly in the browser, at the edge, on-device, or in the cloud. You can read more on the Moss homepage and in the YC listing above.
The numbers are company-supplied, and they matter
Moss gives several traction figures in the YC job listing: deployed across 100+ countries, 5M+ real-time voice minutes served, supporting enterprises that serve 3,000+ of their own end customers, and 380,000+ package installs. These are Moss-published numbers on a hiring page rather than audited metrics, and Moss does not disclose revenue, headcount beyond the listing context, named customers in the listing, valuation, total funding, or round terms beyond YC backing.
Why this hire is strategically important
Voice agents have less tolerance for infrastructure delay than chat interfaces. A chatbot can hide latency behind a typing indicator; a phone agent that s at the wrong moment feels broken. If Moss’s bet is that retrieval cannot live several hops away from the agent, then the SDK layer becomes the product: the install has to be quick, the runtime has to fit each platform, the bindings have to behave consistently, and the latency claim has to survive real devices. That is a hard engineering problem, and Moss is hiring as if the hard part sits below the demo.