At Esyasoft Technologies Private Ltd., I developed an LLM-powered conversational analytics platform that enabled 100+ users to generate real-time insights, reducing time-to-insight from days to minutes.
I also cut Generative AI inference costs by 90% by benchmarking and migrating workloads from OpenAI GPT to Qwen-72B-Instruct while maintaining output accuracy and latency. I streamlined self-service BI and reporting workflows, reducing manual effort by 70%.
At Esyasoft, I built machine learning systems for energy use cases, including models that identified high-risk consumers with 85% precision and a consumer–transformer connectivity framework with 90%+ accuracy. I also developed time-series forecasting pipelines delivering 2% MAPE demand forecasts.
In personal projects, I built a multi-agent RAG finance assistant and a local-first MCP server for data analytics. I hold an M.Tech from IIT Bombay and have also used Python for socio-economic data analysis during a research internship.
