Anirudh Gangadhar

Ai Engineer @University Health Network

Toronto, ON, CA

Architected end-to-end LLM inference pipeline on H100 VM, utilizing vLLM and FastAPI to process over clinical notes for high-throughput batch extraction of Social Determinants of Health (SDoHs), achieving an F1-score of >0.85 across all tested categories. • Engineered autonomous RAG workflow for LabGPT Chatbot using LanceDB, automating the ingestion and semantic retrieval of 40+ disparate open-source codebases to support complex research and LabOps queries.• Developed \"DocuParser\" and CTgov extraction pipelines, creating an automated, multi-stage OCR-LLM workflow for high-variability clinical PDF scans; successfully optimized token usage and cost for large-scale (10k+ clinical trial) document processing, while maintaining high (>95%) extraction accuracy. • Operationalized full-stack LLMOps, including multi-GPU distributed training, model quantization (4-bit/8-bit), and the containerization of AI services (Docker) to standardize deployment and service scalability. • Managed high-stakes clinical cohorts patients), architecting secure data pipelines that bridge raw clinical SQL databases (Clarity/Caboodle) with LLM-ready formats, while maintaining 100% data integrity and strict PHI security standards throughout the compute lifecycle.