Conversational Enterprise Analytics
Built a natural-language analytics workflow using Claude API, metadata-aware prompting, SQL generation, and Snowflake query execution.
I build practical machine learning and AI systems that turn complex data into measurable outcomes. IIT Kharagpur graduate working across applied ML, agentic AI, LLM applications, semantic retrieval, and inference optimization.
Built a natural-language analytics workflow using Claude API, metadata-aware prompting, SQL generation, and Snowflake query execution.
Created an LLM-based document workflow to extract and organize unstructured information into structured Excel deliverables.
Developed a PMJAY investigation model to help prioritize suspicious cases and support fraud-loss reduction.
Current work on embedding-model fine-tuning for better semantic representation and retrieval in regulatory compliance mapping.
Exploring caching and inference optimizations including KV cache, prefix caching, and prefill/decode efficiency.
Applied DeepAR and time-series forecasting methods to help forecast contact volumes and support staffing decisions.
Developed renewal propensity models to support targeted customer engagement and retention in motor insurance.
Built an XGBoost pricing model to support risk-aware pricing decisions and improve revenue outcomes.
Fraud-detection agent concept developed for the Bajaj Insurance Hackathon, combining AI workflows with investigation support.
Integrated MSc, Exploration Geophysics · 2016–2021 · GPA 7.95
Pianist for 10 years, vocalist and guitarist. I led an institute music team that won bronze, and I’m interested in Music Information Retrieval and AI-powered music learning.
Open to conversations about applied AI, LLM systems, and meaningful data products.