Principal Software Engineer
Beacon.liCurrent
Hyderabad, Telangana, India
May 2026 – Present7 mos
- Platform engineer on an AI platform that automates enterprise software implementations, building the platform features needed to onboard new clients and extend existing ones
- Work on the copilot chat service, whose agent loop runs tool calls and gathers the context agents need to answer users
- Fixed context overload from large uploads: files are now chunked into the RAG store, and when a user mentions a file, the 15 most relevant chunks are retrieved instead of the whole text
- Built row- and column-aware chunking for tables in the ingestion pipeline, so chunks never mix rows, columns, or empty cells
- Extended the media ingestion pipeline to detect file types and route them to extraction services, adding support for large files, more formats, image extraction, and content size limits
- Built retrieval-augmented generation with Claude models through an internal AI gateway, using LangChain and Qdrant
- Claude
Retrieval-Augmented Generation (RAG)
- LangChain
- Qdrant
- TypeScript
- KoaJS
- Next.js

