At SHRIVE SOLUTIONS LLC, I built domain-specific banking LLM tuning with PEFT, LoRA, and QLoRA, and served LLM endpoints through vLLM and Triton Inference Server. I also built LangGraph multi-agent RAG for FinTech support, with uncertain cases escalated.
At COPANI, I built production RAG systems and fine-tuned foundation models, raising recall from 73% to 93%. I also shipped FastAPI services on Docker and Kubernetes, cutting incidents by 30%.

