Infinigrid, AI for the power grid
ML engineer at Infinigrid. We build AI systems that watch the electrical grid in real time and predict the risks that have been historically hard to see: overloads, equipment degradation, cascading failures before they cascade.
Why this work matters
The electrical grid was designed for a world that doesn't exist anymore. Loads are peakier, generation is more distributed and weather-dependent, and the data infrastructure to react in real time is still being built. Forecasting and risk-prediction at grid scale is one of the few applied-ML problems where the downstream consequences are measured in megawatts, not click-through rate.
What I work on
Production ML on grid-scale time-series telemetry: forecasting (load, generation, and weather-coupled effects), anomaly detection on streaming sensor data, and the engineering that turns a model into a service the grid can depend on, from data pipelines through evaluation, monitoring, and deployment. Specific methods and results are confidential while the product is early.
How I approach it
Grid data is unforgiving. A wrong forecast costs megawatts, not click-through rate, so calibration and the cost of being wrong in each direction matter as much as raw accuracy. Most of the work is the unglamorous half: clean data, honest evaluation, and pipelines that keep running when the inputs misbehave.