AI Research Engineer
I’m an AI Research Engineer interested in experimental model architectures, especially generative models and world models. I build and benchmark new approaches, break systems down when they fail, and iterate based on what the experiments show.
I’m a Computer Engineering graduate from Penn State University. A lot of my model research has been done on consumer hardware, where architecture and efficiency choices matter. Outside of AI research, I also build and maintain open-source C++ software used by hundreds of thousands of people.
I built an action-conditioned world model around a sparse MoE DiT and trained it on my own data using a single RTX 4090. I experimented with the architecture and conditioning, then built targeted diagnostics and benchmarks to find failure points and inform the next iteration.
I built an open-source C++ replay system that can record and visualize thousands of Geometry Dash attempts at once in real-time. The replay data is quantized and keyframed to reduce size. Most data stays on the disk until it is needed. Time-windowed indexing keeps the system from checking attempts that cannot matter yet. Player objects are reused and ghosts can skip work when they are not relevant to the current section. I update and maintain this project for over 400k users on Windows, iOS, macOS, Android64, and Android32.
At Penn State University, I train and evaluate computer vision models for an insect-monitoring research project, build tools to make the field pipeline easier to monitor and debug, and develop the downstream workflow for organizing research data. A related research paper is in preparation, so I am keeping unpublished technical details off this site for now.
Hands-on work in model research and applied ML.
I work on the ML and software side of an insect-monitoring research project. My work includes model training and experiments, maintaining and adding features to the pipeline, and organizing and creating software for processing research data.
I work on experimental model architectures in a small independent research group. I implement and benchmark model changes, then build targeted tests when training behavior is unclear.
I worked on AnimateDiff training code for custom video-model experiments. I modified and debugged the training pipeline and helped with multi-GPU compatibility and Linux training issues. I demonstrated progress to the client during weekly meetings.
I'm looking for Research Engineer roles where I can spend most of my time implementing model ideas, running experiments, and iterating on architectures. I work best with room to own technical problems while still having strong researchers and engineers around to challenge assumptions and compare results.