Justo E. Karell, simulation engineer and researcher

For both research and work, I enjoy simulating really difficult things; namely complex fluids, plasmas, and industrial systems (e.g., transportation/logistics). In particular, I use analytic and optimization methods to improve simulations and numerical methods for fluids and plasmas, along with modern AI techniques where applicable. A lot of my projects are about either improving simulations, or reducing the time-to-simulation/solution. More is on my Research page and Google Scholar.

For work, I've mostly been a Data Scientist and Engineering Manager across energy, transportation, and defense. I thrive in roles with either Product Management and/or Research Engineering responsibilities.

In teaching, I love showing people how hard things actually AREN'T. This usually means starting from a definitions-based approach, and strictly building on concepts. For Deep Learning, this tends to mean starting from the definition of a Neural Network (which is just the composition of two nonlinear functions), understanding that layers are just functions, then letting students build, break, and redesign standard architectures into useful models for real-world problems. NLP is just putting our natural language into strict rules, which makes teaching it even more fun, as you can integrate philosophical questions that (ironically) actually matter in real-world software systems, such as "what is a word?" which helps to inform tokenization.

Outside of that I surf, skate, blog about stock market catalysts in Weekly Catalysts, watch a lot of stand up comedy, cook for everyone I share an office with, and make wallets.