Professional Experience

Hands-on technical lead managing a portfolio of concurrent ML and GenAI initiatives, working directly with director and executive leadership, and mentoring junior and senior data scientists and interns. Selected projects:

Work Order Intelligence Platform AI Solution Architect & Lead AI Engineer

• Architected LangGraph agentic workflows that extract Problem-Cause-Remedy diagnostics from unstructured HVAC and refrigeration work orders, and use agentic RAG over BigQuery vector search to match assets and components to the asset taxonomy and spec sheets. The LLM-as-judge evaluation pipeline exposed a 54% upstream data-quality gap that drove data-capture fixes.
• Built classification of all work orders into category-specific cost buckets at 98% accuracy, cutting manual data entry by 99%. A JEV model (Typesafe AI) is the primary classifier, with an LLM fallback below 0.9 probability and an evaluation pipeline that scores five criteria to ship, review, or reject each result.
• Designed a proof of concept for a multimodal (text and image) intake agent to replace form-based maintenance requests. It collects missing details, triages, and hands off each request with similar past incidents, likely duplicates, and repair documents; projected to cut time to repair by nearly 50% and cost by 25% by eliminating duplicate work.

Corporate-Card Expense Classification AI Solution Architect & Lead AI Engineer

• Architected and led a LangGraph agentic workflow on Vertex AI that replaced manual expense coding on regulated PII data. It discovered a governed 44-category taxonomy and classified 51,518 transactions with 99.76% high confidence (98.4% in one LLM pass), with automatic OpenAI-to-Gemini failover.

Retail Demand Forecasting Platform Lead Data Scientist

• Led a cross-functional team of data scientists and engineers, setting technical direction and MLOps standards for a 22-repository GCP platform that forecasts SKU-store sales and promotional redemption across approximately 350 stores over an 89-week horizon.
• Cut forecast WMAPE roughly in half versus the legacy model, and pipeline runtime by 72%. Built the production models: LightGBM for weekly SKU-store sales and promotional redemption (benchmarked against XGBoost and TFT) and ARIMA+ for daily SKU-store demand, trained on Ray in Vertex AI, served via BigQuery ML, and fed by Dataform feature pipelines.
• Designed the evaluation and explainability layer (leakage-audited backtests by category, promotion, and horizon week; SHAP dashboards for planners), an experimentation platform, and an on-demand Cloud Run forecasting API returning per-promotion forecasts in approximately 25-30 seconds.
• Previously led modeling for the legacy promotional forecasting platform it replaced, and contributed to refactoring that platform from R and Scala to PySpark on AWS, cutting runtime by nearly 70%.

Electronic Fulfillment Center (EFC) Forecasting Architect & Lead Data Scientist

• Architected and built a 21-day multi-horizon forecast of hourly orders and units for every EFC and store pair (9 EFCs, 73 stores), ensembling an ARIMA baseline, global LightGBM models, and a Temporal Fusion Transformer with extensive feature engineering. It is now in staging at approximately 14% WMAPE.
• Engineered its MLOps pipeline on Databricks (AWS): a multistage workflow with MLflow-tracked experimentation and model selection, weekly retraining, and daily inference via model serving. It feeds EFC staffing and capacity planning (staff allocation partly live), with EFC-to-store allocation optimization and new-EFC planning planned next.

Assortment Space Optimization Lead Data Scientist & MLOps

• Led a platform that decides how much shelf width each product section should get in every store to maximize sales and gross profit. A two-stage hierarchical XGBoost model (segment to store, monotonic constraints) forecasts response curves for every candidate width, including widths a store has never used.
• Formulated a PuLP mixed-integer optimizer that reallocates each store's fixed aisle space to the highest-return sections under merchandising constraints, solved in parallel across stores on Databricks/Spark and delivered through a self-service Streamlit dashboard; projected 2-4% sales and gross-profit lift.
• Rebuilt the data preprocessing, feature engineering, and compute layers, cutting end-to-end runtime by 86% (9 hours to 75 minutes) and compute cost by approximately 90%.

Markdown Optimization Lead Data Scientist / Technical Lead

• Led the price-elasticity forecasting component on Databricks/PySpark: an ensemble of XGBoost, LightGBM, and AdaBoost combined into one forecast with monotonic constraints, trained on four years of store-SKU POS history to produce 38-week demand curves.
• Led the optimization engine: a pruned combinatorial search over multi-week discount schedules that balances revenue against shrinkage and empty-shelf cost, with channel-specific weights across three retail channels, under hard inventory caps and merchant guardrails. It delivers weekly recommendations that merchants typically adopt as-is, with UI overrides.
• Increased markdown revenue by 11% in the most recent benchmark while reducing stock-outs and expired products.

Warehouse Forecasting Solution Architect & MLOps

• Architected and productionized a Databricks/Spark system on AWS that forecasts outbound selection and pallet selection for 12 distribution centers over a 28-day horizon, used to plan warehouse labor and transportation a month ahead with weekly updates. It uses per-series XGBoost models with MLflow tracking, weekly retraining, and daily scoring.
• Built Spark-native accuracy monitoring over 2.2M forecast-actual pairs, holding 9.05% WMAPE on the primary selection forecast flat across the 28-day horizon, and shipped v2 as a zero-downtime shadow deployment.

• Applied GenAI & Agentic AI Teaching: Taught 250+ students to date in Agentic AI and Applied Generative AI certificate programs, with a 4.92/5 instructor rating. Topics cover production LLM applications end to end: prompt optimization (DSPy, GEPA), RAG, GraphRAG, and agentic RAG; fine-tuning; multi-agent systems (LangGraph, CrewAI) with MCP and A2A interoperability; and evaluation, tracing, and guardrails (DeepEval, LangSmith).

• Advisory Board: Serve as AI and technology advisor to a recovery-science wellness startup, advising the founding team on technology strategy, platform architecture, and AI/ML strategy across the product.

• Quantitative Risk Modeling: Developed quantitative risk models using game theory (Nash and Quantal Response Equilibrium) and Monte Carlo simulation to produce risk scores, loss distributions, and control ROI, applying them to cybersecurity in manufacturing systems.
• Cybersecurity Research: Conducted doctoral research on cyber-physical security in manufacturing, using a game-theoretic approach to evaluate and strengthen security measures and data protection. The work was published in peer-reviewed venues (see Publications and Research).
• Operational Efficiency: Improved shop floor scheduling efficiency in a virtual manufacturing environment by applying optimization algorithms, an early example of using data science methods for operational improvement.
• Educational Research: Enhanced predictive accuracy in educational research using statistical and machine learning techniques (regression, classification, clustering), influencing student success strategies.

• Predictive Maintenance: Developed, validated, and deployed an ML-based anomaly detection system for predictive maintenance using supervised and unsupervised techniques, improving operational efficiency and equipment lifespan, reducing downtime by 30%, and saving $200K+ annually.
• Data Infrastructure Development: Designed and established data structures for operations tracking, boosting data analysis productivity and supporting better-informed decisions.

• Maintenance System Optimization: Led implementation of an optimized maintenance system that reduced unplanned downtime, improving machinery health forecasting with statistical and ML models.
• Emergency Maintenance Data Analysis: Applied statistical analysis to emergency maintenance data, reducing emergency repairs and showing the value of data-driven insight for failure prevention.