AI research SaaS: $1.2M seed round closed 2 weeks post-launch

The situation

• The founding team, an ex-Google researcher and a product designer, had created genuinely beautiful UI in v0. Clean, intuitive, exactly right for enterprise research teams. But there was no backend.
• Document processing was fake, it just displayed uploaded files without actually parsing them. Search was client-side filtering of titles only. AI features were non-functional stubs.
• They had a term sheet contingent on a working demo. The investor had set a 3-week deadline. They needed: vector search across 100,000+ research documents, multi-model AI orchestration (different models for different query types), team workspaces with role-based access, and a usage analytics dashboard for the enterprise tier.

What we did

• Document ingestion pipeline: PDF parsing → chunking → embedding → Pinecone upsert
• Hybrid search: vector similarity + BM25 keyword search with Reciprocal Rank Fusion
• Multi-model AI router: Claude for reasoning, GPT-4 for extraction, Mistral for lookups
• Team workspaces with role-based access, shared libraries, and search history
• Enterprise analytics: usage patterns, cost tracking, productivity metrics
• SSO via Google Workspace OAuth for enterprise procurement compliance
• Async document processing queue via Bull + Redis, zero UI blocking

The result

• The seed round closed 16 days after launch. The investor's term sheet was unconditional, no further technical conditions.
• Three months post-launch, the platform has 7 enterprise clients (research institutions and consulting firms), $28K MRR, and a Series A process underway. The founder reports that 100% of enterprise deals cite the analytics dashboard and SSO as table-stakes features during procurement.