ERCOT Net-Load Forecasting Model

AI is reshaping every major industry, and energy is one of the most obvious and important. I was curious about what applied ML actually looks like in a domain with real infrastructure stakes, so I picked a classic forecasting problem and built a real solution.

A gradient-boosted forecasting pipeline for net load prediction on the ERCOT grid, incorporating weather features, calendar effects, and historical demand. Trained on multiple years of public ERCOT data with cross-validated evaluation and an automated retraining workflow.

Python XGBoost FastAPI Docker GCP

Distributed Job Queue

I wanted genuine hands-on experience with foundational systems that underpin modern AI infrastructure. I identified a classic problem in distributed computing, built a clean solution, and learned more from that than from any tutorial.

A distributed task queue built from scratch with worker processes, a coordinator node, and a persistent broker. Supports job prioritization, retries, dead letter handling, and observability hooks. Designed as a learning project for understanding the primitives beneath tools like Celery and RQ.

Python FastAPI Docker Redis Git