Vismay Patel

About

ML/GenAI practitioner with healthcare domain depth who built and deployed a production cervical-cancer EMR classification system using a hybrid rules + medical BERT approach, optimized for high recall under severe class imbalance and PHI constraints. Experienced running end-to-end production ML/LLM pipelines with Apache Airflow (validation, promotion/rollback, monitoring, retraining) and partnering closely with clinicians to calibrate thresholds and implement human-in-the-loop review.

Key Strengths

• Deployed clinical NLP system for cervical cancer EMR labeling that saved manual hours/week and improved follow-up flagging for high-risk patients
• Strong approach to imbalanced medical datasets (SMOTE/sampling, cost-sensitive losses, ensemble with rules to reduce false negatives)
• Recall-focused evaluation and clinician-calibrated decision thresholds using real-world prevalence validation cohorts
• Production MLOps orchestration with Apache Airflow (ingestion → training → evaluation → scheduled retraining, SLAs/retries, validation, promotion/rollback)
• Designs reliable AI workflows with modular testing, golden sets/regression tests, guardrails, and human-in-the-loop controls
• Cross-functional collaboration with clinicians/pathologists/lab teams to align model outputs with clinical reasoning and operational workflows

Experience

Education

Awards

• President’s Leadership Fellowship – San Francisco State University
• Top-scoring graduate project: Litlister – marketplace platform with full-stack architecture (Express, Socket.io, AWS)

Skills

Languages