RAG platform for private equity due diligence

Investment teams in private equity spend a large share of their time reviewing information memoranda by hand. For Silver Investment Partners I designed and built an analysis platform that processes these documents automatically and extracts the relevant company metrics in a structured form.

At the core is a RAG pipeline: documents are ingested from Box.com, chunked, and stored in a Qdrant vector database. Language models are connected through OpenRouter. Financial and operational metrics become comparable across targets and are linked to the firm's investment criteria. I measured the pipeline's answer quality and latencies through Langfuse traces and improved them iteratively.

What I built

• Python-based ingest pipeline: chunking and embedding of the information memoranda
• Semantic search across the entire document corpus
• Next.js frontend for research and evaluation, deployed via Docker

Result: analysis time dropped by roughly 65%. The investment team evaluates potential targets considerably faster and spends the time it gains on actual analysis rather than document handling.