Pittsburgh, PA

Building the anomaly-detection pipeline behind GridWatch under a DOE-funded grid-resilience grant.

• Reduced false-positive rate 40% by switching from a fixed threshold to a learned seasonal baseline

Python · PyTorch

Remote

Summer internship on the forecasting team — built and shipped one demand-forecasting model to production.

• Model shipped to production, cut manual forecast-review time by ~5 hours/week

Python · scikit-learn · Airflow

Pittsburgh, PA

Built dashboards for occupancy and maintenance-request trends.

SQL · Tableau

A streaming anomaly-detection pipeline for substation sensor data — flags likely faults 8–15 minutes before a threshold-based alarm would fire.

• 92% precision at a 10-minute early-warning window on held-out substation data
• Presented at the department's spring research symposium

Python · PyTorch · Kafka · InfluxDB

Source

A small tool that pulls a city's public transit GTFS feed and scores routes by realistic door-to-door time, not just scheduled time.

Python · PostgreSQL · GTFS

Source