AI / Retrieval · Case study

Tickline

Client: A fintech research deskDiscipline: AI / RetrievalStatus: Live demo

Plain-English answers across dense financial filings, with every figure traced back to its source.

The problem

Analysts lose hours digging through 10-K filings for a single number, and they can't trust an answer they can't trace. Our client wanted plain-English answers grounded in the actual document, not a confident hallucination.

Our approach

We built a retrieval-augmented assistant over a working corpus of a public-company 10-K. Every answer cites the exact section, fiscal year, and figure it came from. It runs on hybrid retrieval (vector plus keyword) over pgvector, so the right passage surfaces every time.

What we shipped

A grounded Q&A interface with section-level citations and verbatim financial figures, backed by Neon pgvector hybrid retrieval.

  • Plain-English Q&A over filings
  • Section, year & figure citations
  • Neon pgvector hybrid retrieval
  • LangChain ingestion pipeline
Next.jsLangChain.jsNeon pgvectorOpenAI

What we took away

Trust in a RAG product is won or lost on citations. Surfacing the exact source beside every answer mattered more than any single retrieval trick.

See Tickline in action

This is a real, working build. Open it and try it yourself.

Visit the live demo
Want something like this?

Let's build your version.

Tell us the problem. We'll tell you how we'd ship it.