Resume
查看中文简历
Download My Resume
Professional Summary
Data Scientist with experience building end-to-end machine learning and GenAI systems across healthcare, medical imaging, and applied AI products. Skilled in translating ambiguous problem statements into production-ready pipelines spanning data ingestion, modeling, retrieval, and human-in-the-loop workflows. Hands-on experience with LLM-based RAG systems, computer vision for medical and document imaging, and scalable ML infrastructure in regulated environments. Currently focused on robust GenAI system design, evaluation, and deployment in real-world settings.
Core Skills
Machine Learning & GenAI
- Classical ML: classification, ranking, feature engineering, model evaluation
- GenAI systems: Retrieval-Augmented Generation (RAG), prompt engineering, LLM evaluation
- Embedding-based retrieval, cross-encoder reranking, uncertainty handling
- Human-in-the-loop ML and AI-assisted review workflows
Data & System Engineering
- End-to-end ML / GenAI pipeline design with quality gating and fallback strategies
- Noisy document understanding: PDF parsing, OCR, table extraction, image-based reasoning
- Failure-aware system design for malformed inputs, encoding issues, and partial outputs
- Databricks-based ML workflows and feature pipelines
- Containerized deployment with Docker
- Azure-based ML / LLM platforms in regulated environments
- Cloud-native data workflows and monitoring
- Python, Pandas, NumPy, scikit-learn
- FAISS, Hugging Face (E5 embeddings, rerankers)
- OpenAI / Azure OpenAI APIs
- Kedro, Streamlit
Professional Experience
Johnson & Johnson | Data Scientist (Contractor)
Center of Excellence, Beijing, China | Apr 2024 – Present
System-oriented Data Scientist working on ML and GenAI systems for commercial and medical use cases under strict compliance and data-quality constraints.
Commercial ML Systems for Drug Launch & HCP Targeting
- Built and maintained ML models for early adopter identification in new drug launches using historical sales, promotion, and HCP interaction data
- Designed feature engineering and ranking pipelines to support large-scale HCP prioritization under business and compliance constraints
- Implemented modular data pipelines using Kedro, improving validation, reproducibility, and cross-team handoff
GenAI / RAG Systems for Medical Content Governance
- Designed RAG workflows to detect off-label or non-compliant content by comparing localized medical FAQs with regulatory drug labels (Japanese market)
- Led prompt design, retrieval strategy selection, and error analysis for ambiguous translations and partial semantic matches
- Built quality-gated fallback strategies combining rule-based checks, embedding retrieval, and LLM reasoning
- Designed and deployed a Japanese medical RAG system using FAISS + multilingual E5 embeddings, history-aware retrieval, and cross-encoder reranking
- Built robust JP/CN name normalization and fuzzy matching pipelines across writing systems
- Developed a Streamlit-based UI for retrieval tuning, evidence tracing, and multi-turn dialogue logging
Document Understanding & PDF Table Extraction Systems
- Designed a robust pipeline for extracting hospital schedule tables from heterogeneous PDFs (digital, scanned, malformed)
- Integrated Camelot, OCR-based table detection, and LLM-based image reasoning with quality gating and fallback logic
- Built reviewer-facing UI for inspection, annotation, and batch validation
- Implemented failure-aware handling for encoding issues, broken PDFs, OCR degradation, and oversized images
- Participated in migration of ML workflows to Databricks, supporting feature storage and transformation redesign
- Ensured alignment with platform constraints and compliance requirements
Zenni Optical | Data Scientist, AI/ML Department
Beijing, China | Jun 2023 – Jan 2024
Prescription Understanding & Recognition System
- Designed and deployed a prescription (Rx) recognition service extracting structured optical parameters from user-uploaded images
- Built a FastAPI-based inference service integrating OCR outputs with rule-based and heuristic post-processing
- Adapted parsing logic for Japanese prescription formats, improving accuracy across multilingual and layout-diverse inputs
- Integrated CloudSQL and BigQuery for usage tracking and monitoring
- Investigated image quality issues affecting OCR and extraction reliability
- Analyzed API logs and user behavior to identify degradation patterns
- Built an internal Streamlit-based review tool to support error analysis and manual validation
University of Texas Southwestern Medical Center | Data Scientist
Quantitative Biomedical Research Center
Dallas, TX, USA | Sep 2021 – May 2023
Integrated Imaging & Biomedical Data Systems
- Designed and implemented an image analysis system integrating spatial imaging data with single-cell CyTOF expression profiles
- Built a Flask-based service with parallelized processing and Dockerized deployment
- Achieved ~10× runtime speed-up over baseline pipelines
Cell Segmentation & Classification in Digital Pathology
- Developed a Mask R-CNN pipeline for nuclei segmentation and 6-class classification in H&E pathology slides
- Customized loss functions to handle partially labeled datasets, recovering ~20% of training data
- Achieved 82.5% detection and 82.0% classification accuracy
Clinical Text Processing Support (Supporting Role)
- Assisted with integrating existing EHR de-identification and preprocessing tools into research workflows
Donald Danforth Plant Science Center | Data Science Researcher
Data Science Facility
St. Louis, MO, USA | Feb 2020 – Sep 2021
- Built automated pipelines for RGB, thermal, and hyperspectral plant imaging data
- Developed instance-level segmentation and temporal tracking workflows for leaf growth analysis
- Contributed to the open-source PlantCV project (modules, tests, documentation, training)
- Supported interdisciplinary teams with reproducible analysis and visualization
University of Florida Academic Health Center | Data Science Intern
Precision and Intelligent Systems in Medicine Partnership Lab
Gainesville, FL, USA | May 2019 – Aug 2019
- Processed patient vital-sign time-series data for early risk modeling
- Engineered 24-hour post-admission feature sets
- Reproduced interpolation-based models for irregular clinical time-series
Personal Projects
AI News Agent – LLM-based News Summarization & QA
- Designed and developed an LLM-based AI news agent for daily aggregation, summarization, and question answering
- Built a lightweight RAG pipeline combining web-scraped sources, local embedding models, and OpenAI APIs
- Implemented scheduled crawling, automatic topic tagging, and category-based summarization
- Enabled user question answering over the news corpus via hybrid retrieval
- Developed Streamlit front-end; deployed on Hugging Face Spaces; explored Docker-based deployment
Education
University of Florida
M.S. Electrical & Computer Engineering (GPA: 3.86/4), 2019
Thesis: Switchgrass Genotype Classification using Hyperspectral Imagery
M.S. Industrial & Systems Engineering (GPA: 3.87/4), 2017
Tongji University
- B.S. Physics (GPA: 4.45/5), 2015
Thesis: Correction of Intensity Unevenness in X-Ray KB Imaging
Professional Strengths
- System Thinker: Focus on robustness, failure handling, and end-to-end design
- Fast Learner: Rapidly onboard to new domains and technologies
- Communicator: Translate complex systems for technical and non-technical audiences
- Cross-cultural Collaboration: Bilingual in English and Chinese with US–China experience
Publications
Sheng H, Gutierrez J, Schuhl H, Murphy KM, Acosta-Gamboa L, Gehan M, Fahlgren N. Increasing the Throughput of Annotation Tasks Across Scales of Plant Phenotyping Experiments. Authorea Preprints. 2023 Oct 19.
Rong R, Sheng H, Jin KW, Wu F, Luo D, Wen Z, Tang C, Yang DM, Jia L, Amgad M, Cooper LA. A Deep Learning Approach for Histology-Based Nucleus Segmentation and Tumor Microenvironment Characterization. Modern Pathology. 2023 Aug 1;36(8):100196.
Panda K, Mohanasundaram B, Gutierrez J, McLain L, Castillo SE, Sheng H, Casto A, Gratacós G, Chakrabarti A, Fahlgren N, Pandey S. The plant response to high CO2 levels is heritable and orchestrated by DNA methylation. New Phytologist. 2023 Jun;238(6):2427-39.
"Yu, G., Zare, A., Sheng, H., Matamala, R., Reyes-Cabrera, J., Fritschi, F.B. and Juenger, T.E., 2020. Root identification in minirhizotron imagery with multiple instance learning. Machine Vision and Applications, 31, pp.1-13."
Service and leadership
- Co-Chair of the Committee for Scientific Training and Mentoring at Donald Danforth Plant Science Center
Certificates, Honors, and Rewards
- 1st Place of the “Swarm Behavior on the Grid” track in the Siemens “Tech for Sustainability Campaign 2023” (2023)
- Google Data Analytics Certificate– a rigorous, hands-on program that covers the entire scope of the data analysis process
- 2nd Prize Tongji University Undergraduate Innovation Program (2014)
- National Endeavor Scholarship (2014)
- 2nd Prize, Tongji University Scholarship of Excellence (2014)
