Machine learning engineer with a background in computer science, focused on building ML systems and understanding the models underneath them.
I am a Machine Learning Engineer with a master’s degree in computer science from Purdue University. My focus and passion lies in machine learning; specifically designing and implementing state of the art architectures and custom models.
I make an effort to stay at the cutting edge of ML research, especially fields such as LLM architecture advances, reinforcement learning, computer vision, and natural language processing. I like working close to the technical details and understanding why a model behaves the way it does. I use personal projects to dig into model specifics and learn exactly how they work and how they can be improved.
Professionally, that has translated into experience building software, designing experiments, working with data pipelines, and researching how ML components behave in real systems. My recent work at Kensho has centered on domain specific retrieval-augmented generation and evaluation on production-level tool calling systems.
Investigated grokking: the delayed transition from memorization to generalization in a neural network.
Designed and trained a 124M-parameter Transformer language model from scratch, then specialized it for baseball question answering.
Implemented diffusion models from scratch and trained controlled experiments varying quantization and the number of timesteps.
Explored knowledge distillation for Llama 3.1 1B to create a smaller, capable language model.
Graduate coursework spanning natural language processing, machine learning systems, statistical machine learning, reinforcement learning, computer vision with deep learning, graduate algorithms, parallel computing, and distributed systems.
Undergraduate coursework covering data mining and machine learning, data structures and algorithms, linear algebra, and probability.