Building forecasting systems for global hardware supply chains
Building systems where AI sees around corners.
I work across machine learning, forecasting, computational neuroscience, and technical products. I like building systems that expose hidden structure: opportunity access, reasoning failure modes, medical imaging signals, and now cascading risk in global hardware supply chains. Currently building n2, a forecasting engine for disruptions before they become obvious.
Currently building
/01A forecasting engine for cascading hardware supply chain disruptions.
Building a dynamic dependency graph of suppliers, components, ports, and energy infrastructure. A multi-agent pipeline ingests unstructured signals from local-language news, maritime AIS traces, customs data, grid anomalies, and weather events, then runs probabilistic simulations to estimate downstream disruption risk weeks before it is visible.
dependency graphs
multi-agent ingestion
event extraction
probabilistic simulation
forecast calibration
Builds & projects
/02Pathspire
15k+ users · 20+ countries · solo built
Research and career opportunity platform for students without strong networks. Built, launched, and rebuilt through user feedback, improving matching from scattered discovery to a more systematic opportunity layer.
productmatchinggrowth
AIthena Research Group
independent lab · peer reviewed output
Co-founded an independent AI research group producing work in machine learning, computational neuroscience, and AI reasoning without university affiliation.
MLresearchAI evaluation
Alzheimer's MRI Subtyping
CVPR 2026 · first author
Built a multimodal unsupervised discovery pipeline to identify hidden structural and functional subgroups in Alzheimer's MRI data, testing whether standard diagnostic labels hide meaningful variation.
PythonMRIclustering
Quantum Inspired ML
Google Quantum AI · AIMLSystems oral
Developed and benchmarked quantum inspired tensor decomposition methods for representation learning in high dimensional datasets, leading to a first author oral presentation.
tensor methodsrepresentation learning
MRI Reconstruction Failure Modes
PAI 2026 · Stanford
Built a controlled measurement space analysis showing how corruption structure determines different failure modes in MRI reconstruction.
medical imagingk-spaceevaluation
n2
in progress · supply chain forecasting
Forecasting engine for cascading disruptions in global hardware supply chains, using real-time signals, hidden dependency graph construction, probabilistic propagation, and historical backtesting against rare events.
graphsforecastingagents
Research output
/03CVPR 2026 · First author
Multimodal discovery of hidden Alzheimer's disease subtypes from MRI data.
Built an unsupervised ML pipeline for identifying structural and functional subgroups in Alzheimer's disease imaging data.
ACM AIMLSystems 2025 · First author oral
Quantum inspired tensor decomposition for representation learning.
Research conducted with Google Quantum AI on high dimensional representation learning using tensor methods.
ICML 2026 Workshops · Co author
AI reasoning, epistemic intelligence, and causal faithfulness.
Co authored work accepted to ICML workshop venues exploring reasoning failures, epistemic structure, and model trustworthiness.
CCN 2026 · NYU
Computational cognitive neuroscience research accepted for presentation.
Work connecting machine learning, cognition, and evaluation of reasoning systems.
PAI 2026 · Stanford
Corruption structure determines failure mode in MRI reconstruction.
Controlled measurement space analysis of how corruption patterns produce distinct reconstruction failures.
Timeline
/042026
BuildStarted building n2, a forecasting engine for cascading hardware supply chain disruptions.
2026
AcceptFirst author paper accepted at CVPR 2026 workshop on subtle visual computing.
2026
PresentResearch accepted to PAI 2026 at Stanford and CCN 2026 at NYU.
2026
SelectAdmitted to YC Startup School 2026.
2025
PresentFirst author oral at ACM AIMLSystems from Google Quantum AI research.
2024
ResearchStarted Alzheimer's disease modeling work at Northwestern University.
2023
ShipLaunched Pathspire to help students find research and career opportunities.
Recognition
/05- CVPR 2026, First Author2026
- ACM AIMLSystems, First Author Oral2025
- ICML Workshop Acceptances2026
- CCN 2026, Accepted Presentation2026
- PAI 2026 at Stanford, Accepted Poster2026
- YC Startup School2026
- Congressional App Challenge Winner2024
- ISEF Semifinalist2025
- JSHS State Finalist2025
- PVSA Gold Recipient2025