SEBI Search — Regulatory RAG Platform
Jun 1, 2025
·
1 min read
Project Details
01
Document Pipeline
Built a full RAG platform over thousands of SEBI legal and regulatory documents.
02
PDF Understanding
Parsed PDF files with Docling and created heading-aware document chunks.
03
Semantic Retrieval
Generated Jina v5 embeddings and stored them in Milvus for vector search.
04
Hybrid Search
Combined Jina semantic retrieval with BM25 keyword search for stronger recall.
05
Reranking Quality
Added ModernBERT cross-encoder reranking, reaching an 80% retrieval hit rate and 75.8% nDCG.
06
Grounded Generation
Implemented citation-constrained generation with vLLM/OpenAI, grounding answers to exact file, page, and order metadata.
07
Production Reliability
Added Redis and FAISS caching, Prometheus metrics, health probes, concurrency controls, and request timeouts.
08
Offline Deployment
Delivered 100% offline Docker Compose deployment for restricted SEBI environments on RHEL 9 air-gapped VMs.
Tech Stack
Retrieval & NLP
Docling
Jina v5
Embeddings
BM25
ModernBERT
Reranker
vLLM
OpenAI APIs
Hybrid Search
Platform & Operations
Python
FastAPI
Milvus
Redis
FAISS
Prometheus
Docker
RHEL 9
Authors
Lead Engineer — GenAI / Agentic AI & Applied ML
Lead Engineer with 5 years of hands-on experience developing and deploying GenAI, RAG, Agentic AI, and deep learning systems from scratch. I build production-grade AI platforms for realtime voice interaction, enterprise knowledge retrieval, regulatory search, and automated finance analysis. My work spans NLP, autonomous driving, computer vision, and Camera-LiDAR-Radar fusion, with a focus on reliable architecture, model optimization, and scalable data pipelines.
Stack: Python, GCP, GenAI, LLMs, Agents, Docker, FastAPI, Cloud SQL, Google ADK, Vertex AI, Agent Engine, Cloud Run, OpenAI Realtime API, WebSockets, Redis, MongoDB, Milvus, JWT, Docling, BM25, ModernBERT, vLLM, and FAISS.