Papaya
Optimization engine for production AI agent workflows.
The founders
M(
FV
The company
Papaya is an optimization engine for production AI agents. Its SDK ingests live agent traces — prompts, context, tools, outcomes — and runs over 200 research-backed analyses to flag issues like context bloat, redundant retries and inefficient model routing, then generates pull requests with concrete fixes.
Teams are shipping AI agents to production faster than they can tune them, and the resulting context bloat, excessive tool calls and sloppy model routing quietly burn latency, quality and budget. Observability tools show the mess; Papaya proposes to fix it.
Founded in 2026 and a graduate of Y Combinator's Fall 2026 batch, the company set out to tackle a problem it says engineers hit repeatedly: agents that work in a demo but degrade and overspend once they're live.
The three founders pair go-to-market and operations experience from SafeBase (plus an HBS MBA) with deep technical chops from Amazon — one a Senior Software Development Engineer on large-scale ML systems, the other a Senior Data Engineer on exabyte-scale data platforms.
Who else is building in this space
Funded · $500K
AI agents predicting experiment outcomes in minutes.
Funded · $125K
AI-powered discovery and capture for government contracts.
Funded · $125K
Unblockable browser infrastructure for agentic automation.
Funded · $125K
Private foundation models for financial transaction data.
Thousands more founded every month.
Dealroom tracks new companies as they appear — founders, funding and signals, over the API or in the platform.