I am an AI scientist focused on understanding and interfacing with biological systems. I develop novel neural network architectures and apply them to complex spatial data in vision and biology. In my PhD, I developed neural networks that predict the structure, function, and development of the brain's visual system. I've also worked as the founding Principal AI Scientist at a stealth research startup building flexible and queryable self-supervised learning systems.
Professional Experience
Vice President, AI Research
2026 - Present
- Responsible for the direction and execution of Noetik's AI research agenda, leading a team of exceptional scientists
Director, Machine Learning Research
2026
- Led training and analysis of TARIO-2, a multimodal model trained to predict spatial transcriptomics from H&E images
- Oversaw application of internal foundation models to predict response to drug treatment from pre-treatment clinical samples
- Created internal tools integrating visualization of model inference with internal agentic AI co-scientists
Principal Machine Learning Scientist
2023 - 2025
- Promoted twice, to Senior then Principal, in under 1.5 years
- Designed and wrote from scratch a flexible, scalable ML framework for distributed model training with PyTorch, Ray, and a custom train loop that underlies all ML work at Noetik
- Developed OCTO, OCTO-vc, and TARIO, a family of novel and proprietary multimodal models trained on large-scale patient data. Applied these models to generate insights into patient stratification and target discovery
- Led AI interpretability work, including the design and training of hierarchical and counterfactual sparse autoencoders (SAEs)
- Founding engineer; developed experimental self-supervised ML systems alongside full-stack web applications for interfacing with trained models.
- Invented topographic deep artificial neural networks (TDANNs), the first models to predict the functional organization of visual cortex by discovering brain-like constraints
- Published 16 papers and preprints in computational neuroscience and machine learning, cited by 900+. Presented at leading conferences while working with profs. Dan Yamins, Kalanit Grill-Spector, and Irving Biederman
- Sole developer of a scalable, cost-effective solution for tracking passengers in airports using a custom ML processing pipeline. Includes face detection, OCR, design and detection of custom 3D-printed barcodes in CT scans, real-time dashboards, and ML-based timeseries clustering
- Ran dev-ops, orchestrated cloud resources, recruited and supervised ML/stats interns, generated reports for Department of Homeland Security, secured funding
Skills
EC2 · Sagemaker · Hyperpod · S3
Education
Dissertation: A Unified Model of the Structure and Function of Primate Visual Cortex
Advisors: Profs. Dan Yamins and Kalanit Grill-Spector
Awards and Hobbies
- Co-author of NVIDIA Best Paper in NeuroAI Award, SVRHM @ NeurIPS 2022
- Grew personal habit-tracker into open-source website where 300+ users share their notes on academic papers. Personally reviewed 200+ papers in neuroscience and ML (1/wk for 4 years)
- Highest GPA in USC class of 2016, 2x USC Best Neuroscience Student, NSF GRFP Winner
- Triathlete, guitarist, trail runner, rock climber, unix + vim enthusiast