Robot learning in dynamical systems on one side. Transformer sequence models on fungal genomes on the other. Industry placement at Accenture overlays both.
Applied ML fellowship at Cornell Tech. Selected from 3000+ applicants. Industry placements and a focus on translating research into deployable systems.
Robot learning, perception, and planning
Building ML models that separate stable and unstable regimes in high-dimensional dynamical systems. Boundary-region learning cuts labeled data requirements by ~60% while holding 97% classification accuracy near regime transition points, where reliability matters most.
Machine learning for genomic analysis
Engineered an ML pipeline over 640M+ DNA bases to predict fungicide resistance in plant disease pathogens. Transformer sequence models learn resistance patterns from large-scale fungal genome data. ~73% held-out accuracy across unseen samples.
Papers and preprints will live here as they land. Both labs are in active research and writeups are in progress.