Matthias Huels

Senior Data Scientist at EnergyHub, New York City

At EnergyHub, I get to work on incredibly exciting challenges in forecasting, control, and modeling of Virtual Power Plants. One of my proudest achievements is inventing Dynamic Load Shaping. It's a system that coordinates distributed energy resources to deliver precise power outputs, rivaling traditional power plants. I developed a custom ML-based model predictive control strategy, and it's been a hit with three of the largest U.S. utilities. You can read more about it in my company blog post and this Canary Media article.

I also helped to modernize our data stack (check out the Snowflake blog post) and developed a simulation environment for electric vehicle charging. This platform allows for virtual testing of managed charging solutions. More on this in the company blog post.

Data Scientist at Hella Aglaia Mobile Vision, Berlin, Germany

At Hella Aglaia, I realized my passion for AI and dove headfirst into the field. This was my first full-time role where I got my hands dirty with real-world projects. I co-led the AI Solutions group, delivering production-grade machine learning systems.

We focused on deep learning-based optical inspection systems for manufacturing sites, automating real-time detection of process anomalies. This improved consistency and production quality significantly.

I also developed anomaly detection for vehicle telemetry data, boosting first-time diagnostic accuracy to 80%. It was an exciting time, scaling these solutions globally and seeing the tangible impact of AI in action. Check out the company press release for more.

R&D Engineer / PhD Researcher at Siemens Energy, Mülheim, Germany

During my time at Siemens Energy, I had the unique opportunity to apply my PhD research directly to real-world engineering challenges. It was incredibly rewarding to witness how my work on non-linear dynamics and optimization could be transformed into practical solutions.

This experience was pivotal in shaping my understanding of AI's transformative potential. Seeing AI-enhanced automated workflows in action for the first time, I fully grasped the disruptive impact AI will have on engineering and beyond.

PhD in Mechanical Engineering from Leibniz University of Hannover

My PhD research explored the world of non-linear dynamics, specifically focusing on friction-damped vibrations in gas turbines. I developed a comprehensive simulation framework that combines resonance analysis with ML-enhanced evolutionary optimization. This innovative approach allowed for autonomous exploration of vast design spaces, finding optimal blade geometries that would be difficult to discover through traditional methods. Working closely with Siemens Energy's R&D team, I implemented these theoretical advances into practical engineering applications.

If you have 38€ to spare and want to practice your German, you can grab a copy of my thesis on Amazon.

M.Sc. in Mechanical Engineering from RWTH Aachen University

Focused on numerical methods and computational mechanics. Spent most of my time writing solvers for partial differential equations and getting excited about the math behind them. This is where I really got hooked on the intersection of computing and engineering.

Visiting Scholar at UC Berkeley

Spent an awesome semester in the Bay Area mostly taking grad courses in computational simulation methods. Got to experience the unique Berkeley vibe and a touch of Silicon Valley energy.

B.Sc. in Mechanical Engineering from RWTH Aachen University

Started out thinking I'd be designing machines, ended up falling in love with simulating them instead. Graduated top 3% of the class, but more importantly, discovered that the future of engineering is computational.

Visiting Scholar at UC Davis

My first taste of the American university system. Took a mix of engineering courses and worked on a cool research project. The California weather wasn't bad either! This experience really opened my eyes to the world.

👶🏻 🔥 TinyTorch - A minimal automatic differentiation library inspired by PyTorch. Implements reverse-mode autodiff with tensor operations, neural network activations, and computational graph visualization. Perfect for understanding how modern deep learning frameworks work under the hood.

🐶 📈 BoundHound - A minimal branch-and-bound optimization solver. Demonstrates solving mixed-integer programming problems using the revised simplex method. Best for exploring the foundations of modern optimization solvers like Gurobi.

🎤 Ursem M., Meadow N., Huels M.
Carbon Reduction 2 Devices - Daily Demand Response Targeting Scope 2 Emissions
Presented a case study with Duke Energy exploring customer interest in reducing scope 2 carbon emissions through daily smart thermostat demand response events. 50th PLMA Conference.

📄 Franklin, W.W., Howerter, S., Huels, M., Hines, P.D.H.
Real-world Experience with Residential Demand Response Baselines. 2023 IEEE Power & Energy Society General Meeting.

🎓 Huels, M.
Der Einfluss geometrischer Auslegungsparameter auf die Dynamik reibungsgedämpfter Turbinenschaufeln. Dissertation, Gottfried Wilhelm Leibniz Universität Hannover. ISBN: 9783959005234.

📄 Huels, M., Panning, L., Wallaschek, J.
Combined airfoil and snubber design optimization of turbine blades with respect to friction damping. ASME Journal of Turbomachinery.

📄 Huels, M., Panning, L., Wallaschek, J.
Influence of geometric design parameters on vibratory response and HCF safety for turbine blades with friction damper. ASME Journal of Engineering for Gas Turbines and Power.

📄 Little, J.H., Kauffman, J.L., Huels, M.
Academic Blade Geometries for Baseline Comparisons of Industry-Specific Forced Response Predictions. ASME Turbo Expo 2017: Turbomachinery Technical Conference and Exposition.