- Founded
- 2022
- Headquarters
- San Francisco, CA
- Employees
- 1-10
- Latest Round
- Series A
- Est. Valuation
- ~$150M
Investment Thesis
Relace builds auxiliary coding models and infrastructure designed to make AI code generation faster, cheaper, and more reliable. Their product suite includes Apply models, Embedding models, and Reranking models purpose-built for agentic workflows.
Despite being a small eight-person team, Relace is in production with companies like Lovable, Magic Patterns, and Codebuff. Raised $23M Series A led by a16z.
AI agents can generate code, but merging it into real codebases, maintaining context across files, and deploying at production scale remains fragile and painfully slow.
Small, specialized Apply/Embed/Rerank models on optimized infrastructure that merge AI-generated code at 10,000+ tokens per second and surface codebase context in 1-2 seconds.
Every new AI capability ships as code; 40+ prompt-to-app companies need production-grade coding infrastructure purpose-built for agents rather than human developers.
Models called tens of millions of times by coding agents; in production with Figma, Lovable, Magic Patterns, Codebuff, and Tempo Labs despite an 8-person team.
Preston Zhou (CEO, physics-to-ML) and Eitan Borgnia (COO, Caltech math, UChicago ML PhD) — rare blend of deep research chops and relentless infrastructure execution.