Sakshat Rapkar

AI ML software development as my day to day work building from scratch and integrating AI intelligence and machine learning prediction in software

AI / ML Engineer, TRIARQ Health, Pune, India Mar 2024 – Present

• Designed and deployed a production multi-domain NL2SQL analytics chatbot that lets business users query KPI, HR and

Revenue Cycle Management (RCM) data in plain English, eliminating manual SQL reporting.

• Built and deployed Cloud Agents in Cursor AI to automate development workflows and accelerate engineering delivery.
• Built a domain routing layer combining keyword matching, embedding-based classification and LLM fallback.
• Engineered hybrid RAG retrieval using dense and sparse vectors with Reciprocal Rank Fusion (RRF) in Qdrant for accurate few-shot context retrieval.
• Implemented NL2SQL generation with the Claude API, few-shot prompting, agentic self-correction and SQL safety validation on BigQuery.
• Built agentic conversational memory for multi-turn follow-up questions and automatic query reformulation.
• Built dbt transformation pipelines on BigQuery (staging, intermediate and mart layers) to power the analytics API.
• Deployed backend services on Google Kubernetes Engine (GKE) and Google Cloud Run with Cloud Build CI/CD.
• Developed a claim denial prediction model (Random Forest, Decision Tree) and an ML-based RCM audit system, deployed via

Flask REST APIs and Streamlit dashboards.

Did my graduation in computer science