Zhengpei Cheng

Data Scientist Ai Engineer @Genentech

San Francisco, CA, US

Built “Literature Extractor”, an LLM-powered RAG pipeline using LlamaIndex + LangChain, auto-extracting structured clinical data from publications; reduced review time by 90% with 95% accuracy.• Developed Medical Education AI Assistant, a multi-agent system leveraging Streamlit, LangChain, LangGraph, FastAPIand AWS Bedrock to automate medical education reporting, cutting process from 6 weeks to 3 days.• Built synthetic datasets from ground-truth labels using Python & LLM-based augmentation, improving training and evaluation robustness with Ragas/DeepEval and vector DBs.• Integrated Redis with LangChain-based ChatGPT pipelines to cache embeddings, inference results, and chat history, reducing redundant API calls and improving latency and scalability in enterprise RAG/agent workflows.• Designed and deployed a custom LLM evaluation framework (accuracy, hallucination, reproducibility) to benchmark RAG + Agent workflows, validated by Medical Affairs.• Delivered enterprise-ready GenAI solutions showcased at Town Halls and widely adopted across Medical, Compliance, and Regulatory departments. Built Literature Extractor, an LLM tool that auto-extracts clinical data, cutting review time by 90% with 95% accuracy.

Weill Cornell Graduate School of Medical Sciences

Master of Science - MS, Population Health Science - Health Informatics

China Pharmaceutical University

Bachelor of Science - BS, Biochemistry and Pharmacology

Georgia Institute of Technology

Master of Science - MS, Computer Science