At Key Predict, I led end-to-end development of a test recommendation system combining an LLM with a rule engine, improving the relevance and accuracy of suggested tests. I also replaced legacy NER pipelines with LLM-based CV parsing and developed REST APIs for AI features.
At WonderBiz Technologies, working with client Schneider Electric, I trained and refined object detection models to recognize symbols in P&IDs and developed an Annoy-based template matching model. My projects include a RAG-based CV and job description matcher, real-time video tracking, and models for bird-call counting and OCR.

