Selected projects.

A look at machine learning work I've shipped — each with a real result attached. Filter by discipline to see ML, NLP, analytics, or computer vision.

01+18% recall Churn prediction model dashboard with cohort retention curves

Churn Sentinel

ML · Retention

A gradient-boosted churn model that flags at-risk accounts two weeks earlier, lifting recall 18 points over the prior baseline.

  • XGBoost
  • Feature Store
  • SHAP
02-63% triage time NLP ticket routing pipeline visualization with topic clusters

TicketRouter NLP

NLP · Support

A transformer classifier that auto-routes support tickets to the right queue, cutting human triage time by nearly two-thirds.

  • Hugging Face
  • BERT
  • FastAPI
03+24% conversion Recommendation system architecture diagram with embeddings

Atlas Recommender

ML · Personalization

A two-tower recommender serving 40M daily impressions, lifting click-through conversion 24% in a controlled A/B test.

  • PyTorch
  • Faiss
  • Redis
040.94 AUC Fraud detection anomaly scores plotted over a time series

FraudGraph

ML · Risk

A graph-based fraud detector scoring transactions in real time at 0.94 AUC, saving an estimated $2.1M annually.

  • GNN
  • Kafka
  • Neo4j
0599.2% accuracy Computer vision defect detection on a manufacturing line

DefectLens Vision

Vision · Manufacturing

A CNN defect detector for a production line, reaching 99.2% accuracy on edge hardware with sub-50ms inference.

  • CNN
  • ONNX
  • OpenCV
06+31% forecast acc. Demand forecasting dashboard with seasonal decomposition charts

Demand Almanac

Analytics · Forecasting

A hierarchical demand-forecasting system that improved accuracy 31% and cut excess inventory across 1,200 SKUs.

  • Prophet
  • LightGBM
  • dbt
07+0.12 ROUGE Document summarization interface highlighting extracted key points

BriefBot Summaries

NLP · Knowledge

A retrieval-augmented summarizer for legal documents, improving ROUGE-L by 0.12 while keeping citations verifiable.

  • RAG
  • LangChain
  • pgvector
08+9pt uplift Causal inference uplift modeling chart comparing treatment groups

Uplift Compass

Analytics · Marketing

A causal uplift model that targets only persuadable users, adding 9 points of campaign ROI while shrinking spend.

  • CausalML
  • DoWhy
  • Snowflake
093x throughput Satellite imagery segmentation map highlighting land use regions

GeoSegment

Vision · Geospatial

A semantic segmentation model for satellite imagery that tripled labeling throughput for a land-use analytics team.

  • U-Net
  • Rasterio
  • Sagemaker
120+Models in production
9yYears experience
3Patents & papers
40MDaily predictions

Have a dataset and a deadline?

I take on a small number of engagements at a time so I can stay hands-on from problem framing to production. If your problem is a fit, let's talk.