At Alphastream Technologies Pvt. Ltd., I build and deploy LLM-powered document intelligence pipelines that turn unstructured documents into structured, validated data. My work uses retrieval, chunking, embeddings, and prompt grounding.
I developed RAG-based extraction workflows for enterprise documents and optimized prompts and retrieval to improve extraction quality while balancing latency and inference cost. One workflow achieved approximately 85% extraction accuracy.
I also developed FastAPI services and REST APIs backed by PostgreSQL, integrating LLM outputs with downstream applications. I work with Claude and Gemini APIs and deploy AI services on AWS Lambda, S3, and EC2.
On AlphaMCP, I contributed to an agentic AI chatbot platform, developing section-specific MCP tools for query-intent-based tool selection. I also developed a document extraction pipeline for deal summaries and reference data from tax notice documents.
