Abdul Mohammed

About

Built and productionized an LLM-powered clinical documentation and insights pipeline at Cardinal Health using LangChain + GPT-4 with RAG to summarize long clinical notes, extract medication/dosage entities, and generate structured SQL-ready outputs for downstream analytics. Emphasizes clinical reliability via labeled benchmarking (precision/recall/F1), shadow deployments, clinician human-in-the-loop review, and ongoing monitoring/orchestration with Airflow, Lambda, S3, Postgres, and Power BI.

Experience

Education

Key Strengths

• Built and deployed production LLM clinical documentation/insights pipeline used across multiple departments
• Strong hallucination/medical safety mitigation using RAG grounding plus rule-based validation and constrained prompts
• Rigorous evaluation methodology (labeled eval sets; precision/recall/F1 across extraction, summarization, classification)
• Shadow deployment and clinician human-in-the-loop review to validate against real clinical workflows
• End-to-end ML/LLM pipeline orchestration with Airflow (S3 ingestion through retraining to Postgres/Power BI outputs)
• Operational excellence: monitoring drift/latency/errors and improving DAG reliability with sensors, SLAs, retries/backoff
• Pragmatic model selection (fine-tuned BERT for fast extraction; GPT-4 + RAG for grounded long-text reasoning)
• Effective cross-functional collaboration with clinical managers; translates metrics into business impact and prioritizes quick wins