Captain
Accurate knowledge search that scales
Winter 2026ActiveB2BInfrastructureDeveloper ToolsBig DataData EngineeringSan Francisco, CA, USA
Company
https://runcaptain.comCaptain delivers the most accurate general-purpose retrieval engine for unstructured data.
Connect file stores and effortlessly retrieve knowledge with much higher accuracy than RAG (Avg: 78% → 95% + citations).
Verdict
High Signal
Market Opportunity
Knowledge retrieval and enterprise search infrastructure is a massive and fast-growing B2B market as every enterprise scrambles to make internal data AI-queryable. ICP is reasonably clear (companies with large unstructured file stores needing accurate retrieval). API-first infrastructure positioning enables broad monetization via usage-based pricing.
Low Signal
Founder Signal
LinkedIn data is unavailable for both founders, making it impossible to verify work history, years of experience, or prior companies. Edgar Babajanyan lists 'Prev. Purdue' suggesting recent grad status with limited professional track record. Lewis Polansky's background is entirely opaque — no evidence of prior exits, senior roles, or relevant engineering/product experience.
Low Signal
Competition
This space is extremely crowded: LlamaIndex, LangChain, Cohere Embed, Pinecone, Weaviate, and dozens of RAG-focused startups all compete here. OpenAI and Anthropic are building retrieval natively into their products. The differentiation claim (higher accuracy) is meaningful if proven but not yet a durable moat — any well-resourced competitor can claim similar improvements.
Medium Signal
Product
Claims a specific accuracy improvement (78% → 95%) over standard RAG with citations, which is a concrete and testable claim. Website implies a live API product with file store connectors. However, no customer logos, revenue figures, live demo, or named testimonials are present in the data.
OverallC Tier
Captain is attacking a real and large market with a credible technical claim (accuracy improvement over RAG), but almost everything else is a concern. Founder LinkedIn profiles are unavailable, making it impossible to assess whether the team has the depth to execute in an intensely competitive space. The competition is brutal — OpenAI, Cohere, Pinecone, and a dozen YC peers are all building here, and a 17-point accuracy improvement is not a defensible moat without proprietary data or architectural lock-in. No press, no customer logos, and no visible traction data makes this look very early. Could be a B with founder evidence and customer proof, but right now there's not enough signal to justify it.
