At Nusummit Ltd, I built NSE and BSE circular crawlers and designed the metadata ingestion pipeline for an AI-powered enforcement search platform. It brings document crawling, Oracle metadata storage, embedding generation, and Qdrant indexing together.
I integrated Elasticsearch BM25 search with Qdrant vector search to enable hybrid retrieval across enforcement cases, circulars, and inspection records. I also generated embeddings with sentence-transformer models for semantic search over regulatory documents.
For the Announcement Search platform, I built a company-matching module to resolve trading-member and company identities during ingestion. I engineered prompt templates and local Ollama LLM interaction logic for grounded responses to natural-language queries over announcements and news.
Before Nusummit Ltd, I worked at NSEIT Ltd on Oracle SQL/PL-SQL development and Python reporting automation. I automated reporting processes end to end, improving turnaround time and reporting accuracy by 30%.
