About Me
Hi there!
I am currently a Member of Technical Staff at Argmax where we’re working on on-device foundation model inference. Previously, I also worked at Groundlight AI and ThirdAI where I worked on a variety of machine learning and engineering problems. Prior to that, I was a student at The University of Chicago where I double-majored in Computer Science and Computational and Applied Mathematics. I was fortunate enough to be advised by Professor Rebecca Willet.
Among other things, I am also the maintainer of flatnav, a robust and memory-efficient library for performing vector search at scale.
My research interests, broadly defined, are in statistical learning theory and machine learning systems. For the latter, I’m mostly excited about devising fast and hardware efficient algorithms for training and inferencing large language models. This involves leveraging sparsity patterns in both the data and the model. I also enjoy working on efficient algorithms for high dimensional vector search on dense vector embeddings using graph-based techniques.
Publications
(Preprint) Blaise Munyampirwa, Vihan Lakshman, Benjamin Coleman. "Down with the Hierarchy: The 'H' in HNSW stands for 'Hubs'quot;. (https://arxiv.org/pdf/2412.01940)
Deep learning detects actionable molecular and clinical features directly from head/neck squamous cell carcinoma histopathology slides. J. Dolezal, J.N. Kather, S. Kochanny, J. Schulte, A. Patel, B. Munyampirwa, S. Morin, A. Srisuwananukorn, N. Cipriani, D. Basu, A. Pearson. International Journal of Radiation Oncology, Biology, Physics, Volume 106, Issue 5, 1165 (https://www.redjournal.org/article/S0360-3016(19)34202-6/abstract)
Talks and presentations
Optimizing HNSW in the age of vector databases
Talk at Amazon Search, Palo Alto, CA
[Slides]