Zhuohan Yu

About Zhuohan Yu

Data Scientist with 8+ years of experience in machine learning, deep learning, and data engineering. Hand-on experiences on LLM fine-tuning, RAG models, prompt engineering, and UI development. Expertise includes statistical analysis, model evaluation, and end-to-end machine learning pipelines to drive business insights and decision-making. Strong background in healthcare analytics, fraud detection, and marketing optimization with a proven record of cost savings and improved ROI through data-driven strategies.Programming Languages: Python (OpenAI, Streamlit, LangChain, Gradio, HuggingFace, Scikit-learn, NLTK, FAISS), Pyspark, SQL, SAS, RDatabase ecosystem: MySQL, SQL server, Oracle, AWS S3, Hadoop, IBM DB2, SQLite, Redcap, NetezzaTools: Apache Spark, Apache Ambari, TensorFlow, R studio, Jupyter notebook, Rshiny, Dataiku, AWS SageMaker, MobaXterm, Crontab, Jenkins, Colab, DataBricks, Azure storage blob, Jira, Confluence, Collibria, Power BI, Tableau, PineconeStatistics & Methods: Large Language model (LlaMa2, Davenci, Gpt-3.5-turbo, etc.), LLM fine-tuning, transformer, Prompt engineering, Linear & Logistic Regression, Gradient Boosting, Naïve Bayes, K-means, ECOD, LOF, Isolate Forest, HBOS, Random Forest, Decision Tree, XGboost, NLP, BERT, Topic modeling, LDA, recommendation system, Neural Network, A/B testing

Experience

1. Senior Data Scientist

   CVS Health

   Mar 2025 — Present

   Unified Measurement Engine Platform (Architectural Leadership)• System Consolidation: Independently architected a unified measurement database consolidating three fragmented systems (Eppo, Ipsos Universal Control), creating a \'Single Source of Truth\' for marketing performance.• Business Impact: Enabled Finance-approved standardized metrics for $50M+ marketing budget decisions, reducing analyst reconciliation effort by 60% and eliminating $140K in annual operational waste.• Smart Sampling Pipeline: Engineered an automated sampling pipeline for 95M+ records, achieving <1% covariate balance across demographic and behavioral variables, thereby eliminating selection bias and improving statistical power.Bayesian Marketing Mix Model • End-to-End Development: Developed a comprehensive Bayesian pipeline using PyMC to optimize pharmaceutical marketing spend, incorporating hierarchical priors with CPM-based constraints. • Attribution Excellence: Achieved 94% model coverage while quantifying campaign attribution across 7 digital channels with monthly contribution analysis for executive decision-making.Evergreen Campaign Measurement & Marketing Effectiveness• Business Impact: Quantified $20M annual incremental margin with 7.9x ROAS through advanced causal inference framework, deploying Propensity Score Matching and Bayesian Structural Time Series for weekly store-level evaluation across 700+ retail locations, replacing legacy cohort-level analysis and improving attribution accuracy by 25-35%.• Operational Excellence: Built automated measurement pipeline with dynamic treatment period calculation and fiscal week arithmetic, reducing time-to-insight from days to hours while processing 700+ stores with 100% success rate, enabling data-driven resource reallocation for multi-million-dollar weekly marketing investments.

Education

• Northeastern University

  Master of Science, Data Analytic Engineering

• Northeastern University

  Master's degree, Data Analytics Engineering

  2016 — 2019

• University of Southern California, Los Angeles, CA

  Master of Science, Mechanical Engineering

• China University of Petroleum, Beijing

  Bachelor's degree, Thermal energy and power engineering

• University of Southern California

  Master's degree, Mechanical Engineering - Thermodynamics

  2014 — 2016

Skills

• R
• Execl
• Matlab
• Oracle Sql Developer
• Data Analysis
• Drawing
• Autocad&Ip
• Visio
• Interactive Operation Research Tutorial
• Flash
• Sas Programming
• Mysql
• C
• Microsoft Office
• Tableau
• Flash Professional
• Lingo
• Vensim
• Autocad
• Python