Finance Document RAG Knowledge Base (Vellum)

End-to-end RAG knowledge base for equity research, ask questions and get cited answers from filings, news, transcripts, PDFs, notes, and URLs.

RAGFintechVector DatabaseNLP

01
Retrieval Augmented Generation
02
Vector database + hybrid retrieval
03
Natural language processing
04
Python API development

The brief

Equity research means jumping across SEC filings, news, transcripts, and private notes, then grounding every claim in a source. Generic chat tools answer confidently without reliable citations, and they don't mix public filings with a private document vault in one retrieval path.

The outcome

01End-to-end document retrieval and drafting system for equity research Q&A
02Hybrid retrieval (dense + keyword + RRF) with Cohere rerank and Voyage embeddings
03Reusable patterns for confidential document Q&A, private vault plus grounded Claude answers

The build

Built an end-to-end RAG knowledge base for equity research. Users ask questions and get cited answers from filings, news, transcripts, PDFs, notes, and URLs.

Document ingest and chunking feed Postgres metadata storage and Qdrant vectors. Retrieval is hybrid (dense + keyword + RRF) with Cohere rerank and Voyage embeddings. Claude handles routing and grounded answers.

Also shipped a private document vault and a Next.js chat UI, including a Neural Trace graph that shows how a query connects through sources, facts, metrics, and synthesis.

Specs

Platform: Next.js chat UI + Neural Trace graph view for grounded answer inspection
Ingestion: Document ingest and chunking for filings, news, transcripts, PDFs, notes, and URLs
Retrieval: Hybrid dense + keyword + RRF, Cohere rerank, Voyage embeddings
Storage: Postgres metadata + Qdrant vectors + private document vault
Generation: Claude for routing and grounded, cited answers

Stack

AI / ML

Anthropic ClaudeVoyage embeddingsCohere rerank

Retrieval

Hybrid dense + keywordRRF fusionQdrant

Backend

PythonAPI DevelopmentPostgres

Frontend

Next.jsNeural Trace graph UI

Building something in this layer?

Start an inquiry

More AI-first

All AI-first

AI Personalised Kids Video Platform

Personalised animated storytelling for children. Generative video pipelines, parent-safe workflows, and infrastructure tuned for viral traffic and repeat renders.

AI Content Production Platform

Content and media production platform, bulk Excel-driven workflows for AI-generated images, video, and audio, with pause/resume job control.

AI-Native Design Studio

AI-native design studio for interior designers: client project management, redesign generation, and team communication in one platform.