Curriculum Vitae

Conducting research with Prof. Micah Goldblum and the Columbia Core AI Lab. Compute supported by Google TRC and the hillclimb residency.

Co-founded Autumn AI and was accepted into Y Combinator. Initially focused on RWD, prior auth, nursing, and voice AI evals, later moved to people intelligence. Secured large contracts with healthcare partners, trained custom models, and developed a voice AI prior auth pipeline using a HIPAA-compliant AWS stack.

Co-authored a paper on differentiable causal discovery accepted at NeurIPS CALM. Designed novel knowledge graph evaluation methods and trained models for automatic knowledge graphs construction from unstructured data. Worked on diffusion-based generative models for use in drug-discovery pipelines.

Selected as 1 of 30 students to join CERN Openlab (CS R&D). Worked under Prof. Brian Cole (Jan-Apr) to develop predictive models for calorimeter modules at ATLAS. Worked on AtmoRep (May-Sep), a 3.5B-parameter transformer for unsupervised atmospheric representation learning in collaboration with the Jülich Supercomputing Centre.

Revamped inventory tracking, recovering over $1M in missing product. Incorporated DALL-E 2 and text-davinci-002 into products and services. Led the design of database architectures, ML pipelines, and automation scripts, including Amazon SP API integration. Built an API around database-integrated programs and deployed on a VM. Taught a 10 part course on calculus, linear algebra, deep learning, and how to build an LLM from scratch in PyTorch.

Worked under Prof. Ray Lee at the Zuckerman Institute. Built experiment-stimulus patient interfaces for use in fMRI machines. Developed RCNN and LSTM models in TensorFlow to predict voxel-level fMRI activation patterns from visual stimuli. Designed CAD models in SolidWorks for 3D-printed components used in dual-head fMRI holders.