Selected
Projects
S2Y Health
Designed a reusable MLOps framework for forecast model governance including SHAP-based explainability layers, model version control, forecast bias detection, and automated monitoring dashboards. Built to meet executive auditability and regulatory requirements across enterprise environments with full lifecycle documentation.
S2Y Health
Founded and built an evidence-based AI consultation platform for Long COVID recovery. Combines LLM-powered symptom analysis with taVNS neuromodulation and HOCL therapy protocols. Developed full-stack agent framework processing peer-reviewed medical literature to deliver personalized recovery plans to global patients.
CertiK
Built CertiK's entire data/ML organization from 0 to 10. Designed and productionized real-time ML systems — Security Score and Social Sentiment engines — running on TB-scale blockchain data. Owned full pipeline: API ingestion → dbt/Airflow orchestration → feature engineering → model deployment and monitoring.
CertiK
Architected an AI intelligence pipeline combining news analytics, social sentiment scoring, and LLM-driven entity extraction to generate high-value security alerts and insights. Designed multi-source ingestion with structured output for enterprise clients needing real-time market and risk intelligence.
Baidu
Designed and owned end-to-end ads performance analytics at Baidu — one of the world's largest search ad platforms. Built monetization measurement frameworks covering CTR/CVR modeling, bid strategy evaluation, and ad quality scoring. Developed data-driven recommendations that directly shaped product and pricing strategy decisions at executive level.
Led A/B and multivariate experiment frameworks for large-scale consumer products across Baidu, Qunar, and TAL. Built customer segmentation systems using RFM analysis, behavioral clustering, and propensity scoring to drive personalization, retention, and growth. Designed experiment governance pipelines with guardrail metrics and causal inference.
TAL Education
Built production revenue forecasting models using GBDT/XGBoost and time-series methods for TAL Education (NYSE: TAL). Designed rolling forecast frameworks incorporating seasonality, promotional drivers, and operational constraints. Automated data ingestion and financial metric calculations with full variance analysis reporting to business leaders.
Baidu
Built forecasting and KPI systems for Baidu's core monetization products over 5 years. Designed executive dashboards delivering revenue, growth, and operational KPIs to senior leadership. Implemented data validation rules and cross-system consistency checks to ensure near-perfect forecast reliability consumed at board level.
