At FlocCare, I built and deployed NLP pipelines for sentiment analysis, text classification, and LLM-based summarization. DPO and LoRA fine-tuning improved accuracy by up to 28% over baseline.
I also created reusable data pipelines in Python and tracked experiments with MLflow, saving three full-time workdays per week. I containerized quantized models with Docker on Google Cloud, reducing inference compute costs by 29%.
At Tellis Technologies Pvt Ltd, I developed an LLM-powered conversational agent for candidate screening that reduced manual effort by 78%. I also built a multi-agent orchestration system and deployed it with Docker on RunPod, reducing GPU costs by 32%.
My projects include a vLLM inference cluster, an adaptive hybrid GraphRAG system, and an agricultural climate intelligence project using causal inference. I ranked 11th in Amazon ML Challenge 2025 and finished in the top 2% globally in BirdCLEF 2026.
