Classical ML

My Approach

Not everything needs a large language model. A lot of the highest-impact ML work I have done involves classical techniques: time series forecasting, classification, regression, computer vision, and predictive maintenance. These methods are well-understood, interpretable, and often the right tool for the job in industrial and operational settings.

What This Looks Like in Practice

Industrial Optimization

At Rio Tinto, I used ML to optimize titanium smelting furnaces and mine feed operations for a large open-pit mine. I also built predictive maintenance models to diagnose and predict equipment failure.

Computer Vision & Speech Recognition

At Axon, I worked on Automatic License Plate Recognition (ALPR) and Automatic Speech Recognition (ASR) products, with a focus on fairness evaluation across different demographics and locales.

Time Series Forecasting

At Wavo, I developed forecasting systems to predict API call patterns and prioritize expensive external data requests. At Rio Tinto, time series analysis was central to the furnace optimization and equipment health work.