Looking for a job

Mitul Agarwal

@mitulagarwal

Message

I build production LLM systems for clinical automation, retrieval, and high-throughput inference.

India

Message

At Global Health Ltd., I build and deploy LLM systems for a 3,000+ bed hospital chain, including automated ICD-10 coding, prescription digitization, and clinical record retrieval. My ICD-10 system processes 8,500+ monthly discharge summaries and achieved 80% exact-match accuracy on validated records.

I optimize on-premise vLLM serving on 4× RTX 4090 GPUs, sustaining 240 tokens per second generation and 1,300–1,500 tokens per second prompt processing. I also build LLM architectures from scratch in PyTorch and developed a real-time paper-trading platform with FastAPI, PostgreSQL, and WebSockets.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist 1

Current

Global Health Ltd.

Jul 2024 - Present (2 years 2 months)

Automated ICD-10 coding for 8,500+ monthly IPD discharge summaries using RAG and vLLM, achieving 80% exact-match accuracy. Deployed on-premise LLM inference on 4x RTX 4090 with dual-endpoint failover and engineered an OPD prescription digitization pipeline using Qwen3-VL.

RAG VLLM LLM Inference Optimization Prompt Engineering Docker Python Qwen3VL PII Masking Vector Databases Model Serving

Data Science Intern

Global Health Ltd.

Jan 2024 - Jul 2024 (6 months)

Developed a Python and PowerBI Costing & Procurement Dashboard with service-level profitability insights, driving an 8% reduction in operational costs. Streamlined 10+ internal reporting workflows using Python and SQL, cutting data turnaround times from days to under a minute.

Python Power BI SQL Data Analysis Dashboarding ETL