I'm an AI Researcher at Microsoft Research: AI Frontiers

I specialize in the design, architecture, and programmatic generation of high-scale training datasets for frontier multimodal models. My work centers on building automated generation engines, complex schema design, and analytical data mixtures that teach models how to reason.

I own the data engineering lifecycle, schema design, and synthetic generation infrastructure for the foundational Phi-4-reasoning-vision-15B model.  Before that, I worked on the Phi-4 language model.

Gmail address: mah5044

In a past life, I was a Member at the Institute for Advanced Study; before that, I was a postdoc at Carnegie Mellon, MSRI, and Lehigh University.  I completed my Ph.D. at Pennsylvania State University in 2017, advised by Sergei Tabachnikov.

There, I studied a broad range of problems in geometric topology, topological combinatorics, and dynamics: embeddings, coupled embeddings, totally nonparallel immersions, k-regular maps, great circle fibrations of spheres, line fibrations of Euclidean space, the geodesic complexity of configuration spaces of trees, and billiards in various settings, for example on Finsler manifolds and on spaces of geodesics.  See my Research page or CV (updated 10/2022) for detailed explanations!