At BlackHole Infiverse India, I architected and deployed a production fraud detection system that reached 94% classification accuracy and cut enterprise fraud risk by 37%. I also engineered a Go transaction and authentication microservice and developed agentic AI workflows to evaluate LLM hallucinations and factual consistency.
I built asynchronous scraping and ETL pipelines processing over 1,000,000 records a day, and trained an LSTM model for disaster response and time-series forecasting that improved prediction accuracy by 37%. In my projects, I developed a self-correcting RAG agent that reduced retrieval hallucinations by 42% and a fraud-scoring service with sub-20ms inference latency.
