.../Web Application...
EigenTalk
A RAG-powered research assistant that turns uploaded documents into a searchable, conversational knowledge base — built in Next.js and TypeScript.
.../overview...
EigenTalk is a RAG (retrieval-augmented generation) research assistant for document-specific AI analysis — it reads a user's uploaded documents and answers questions grounded in those sources rather than the open web. I built the responsive Next.js + TypeScript frontend and the integration layer connecting it to the RAG pipeline, covering the full path from document upload through to a retrieved, sourced answer.
Problem
Researchers and analysts needed answers grounded in their own documents, not the open web — where a general chatbot will confidently hallucinate. The product had to read uploaded documents and answer questions with retrieval you could actually trust.
Role
Frontend & integration engineer — owned the path from document upload to a retrieved, sourced answer.
Implementation highlights
- 01Built a responsive Next.js + TypeScript frontend with Tailwind CSS, designed around the retrieval loop — ask, retrieve, ground the answer in the source.
- 02Wired the UI to the RAG pipeline through REST APIs for document ingestion and AI retrieval.
- 03Used route-level code splitting to keep initial load fast despite document-heavy views.
- 04Added robust error handling around ingestion and retrieval so a slow model response or failed upload degrades gracefully instead of breaking the session.
Outcomes
- ›Delivered a working document-Q&A tool that turns uploaded files into a searchable, conversational research assistant.
- ›Kept the interface fast and resilient under flaky network and model latency via code splitting and graceful error handling.
- ›Kept answers anchored to the user's own sources rather than generic web data.