About me
⚙️ I build ML systems from raw data to production — no handoff required.
That means: data pipeline design, feature engineering, model training and selection, SHAP explainability, MLflow tracking, scoring pipeline deployment, and monitoring. I’ve done this across electric load forecasting (167M row time series), student survival modeling (NLP-enriched, daily batch scoring), genomic survival analysis (~20K features, 1,300 samples), and agentic trading research systems with automated LEAN back testing loops.
The thread across all of it: I treat ML as an engineering discipline. Models ship, pipelines run, results are reproducible, and metrics are tracked.
Recent technical depth:
🤖 Agentic AI systems — LangChain, tool-calling loops, structured JSON reasoning, multi-step optimization
🗣️ NLP feature engineering — VADER/RoBERTa sentiment, zero-shot distress classification, embedding drift
📈 Time series — LSTM, GRU, RNN, MLP on smart meter data; cyclical feature encoding, seasonal decomposition
🧬 Survival analysis — Cox PH, Gradient Boosting, XGBoost on high-dimensional genomic data
🚀 ML Ops — MLflow, GCP Vertex AI, feature stores, pre-compute pipelines, latency optimization
At Wayfair I built ML platforms at scale — the feature store I launched on Vertex delivered ~$4M in its first year; latency cuts on real-time serving unlocked ~$50M in revenue impact.
James’ energy is infectious and creates a positive working environment for his team.
— Mayank Gandhi, Head of Data & ML Platforms, Wayfair
James is great to work with, skillfully leading all the stakeholders on the journey, and motivating them with his infectious enthusiasm.
— Phil Gurbacki, Head of Product, Weights & Biases
James Sanders 