RAG Systems
Production RAG Pipeline with LangChain & FAISS
Implementing Hybrid Search (Dense + Sparse BM25), Reciprocal Rank Fusion, and RAGAS evaluation.
Agentic AI
Building Stateful AI Agents with LangGraph
Cyclic stateful graphs, memory persistence, tool execution, and Human-in-the-Loop workflows.
Document AI
DocuAI Studio v3.1 Architecture
FastAPI-powered document intelligence platform with async chunk ingestion and vector search.
LLM Architecture
Demystifying Transformers & LLM Attention
Deep mathematical and structural walkthrough of Self-Attention, QKV projections, and multi-head mechanisms.