Cameron Thieme

Machine Learning for Dynamical Systems

Research on the reliability of long-run predictions made by machine learning models

Performed Bayesian analysis to quantify accuracy of insights derived from data-driven dynamics

Applications in fields ranging from cellular biology to robotics.

Main language: Python. Also includes shell scripts for training and analyzing models in parallel on a computing cluster using a Slurm workload manager

Machine Learning Algorithm Implementation

Python implementation of Gaussian process regression with interval arithmetic

Based on Algorithm 2.1 of Gaussian Processes for Machine Learning

Forecast of Gas & Diesel Prices (Continuous Data Integration)

Software automatically predicts US average gas and diesel prices for the coming week

Continuously integrates latest data from The U.S. Energy Information Administration automatically

Predictions made using XGBoost; hyperparameters selected using Bayesian Optimization

AWS: Electronic Container Registries (ECR), Elastic Compute Cloud (EC2), AWS CLI, setting permissions

Other skills: Python, Docker, Makefile, virtual environments, Git

Forecasting Hydroelectric Power

Consulting project for Cargill as part of the 2021 IMA Math-to-Industry Bootcamp

Power levels needed to inform future investment strategies

Data mined from public sources

Time series analysis in Python using SARIMAX, deep learning, and other machine learning techniques

Predicted power production with under 4% error 12 months in advance

Data Engineering and Cloud Computing

Simple ML project to develop software and cloud computing skills

Project implemented in CI/CD framework so that model can be continuously updated with new data

AWS: Electronic Container Registries (ECR), Elastic Compute Cloud (EC2), Simple Storage Service (S3), AWS CLI, setting permissions

Other Software skills: Python, Docker, Makefile, virtual environments, .env, Git