Work

Portfolio

Selected work across industry, doctoral research, and ongoing projects.

Recent Work

Coming soon

I am currently studying LLM post-training and reinforcement learning algorithms. This section will be updated with projects and write-ups as they take shape.

Optimal Dynamics

During five years as Director of AI, I built and led the technical foundations behind many products within Optimal Dynamics' Transportation Decision System, bridging CASTLE Lab research with production-scale freight optimization.

Decision-Native Agents

I pioneered the agentic orchestration layer at Optimal Dynamics. Atlas, the Agentic Automation Layer, turns optimization outputs into coordinated, autonomous action. Rather than reactive task-takers following simple scripts, these decision-native agents are powered by the core optimization engine and understand the full complexity of the network.

Bid Platform

I was the architect of the Bid Platform, which helps truckload carriers respond to RFPs with confidence. The product overlays bid lanes on an optimized digital twin of the carrier’s network to uncover the best lanes, optimal volumes, and new business opportunities.

Time Series Forecasting

I owned the custom-built time series forecasting engine that powers all major Optimal Dynamics products. Distributional forecasting is a key component of stochastic optimization engines, and my work at CASTLE Lab on modeling uncertainty directly informed how we built forecasting systems at scale.

Data Platform

I was the architect of modernizing the company’s data platform, which maps, ingests, and pulls together carrier data and operational systems into a unified decision-making platform. This foundation feeds every product in the Transportation Decision System, spanning data mapping, entity modeling, and lifecycle management from raw ingestion through to the optimization engine.

PhD Research

Multi-Agent Sequential Decision Modeling

My doctoral work focused on designing policies where agents can collaborate to learn effectively while implementing decisions that impact the environment. I built detailed simulators for wildfires and epidemics to design and evaluate control systems for these agents.

We extended Warren Powell’s canonical framework for modeling sequential decision problems from a single-agent model to a multi-agent model, and explored algorithms for inserting tunable parameters into deterministic lookahead policies to control risk.

Stochastic optimization for vaccine and testing kit allocation for the COVID-19 pandemicEuropean Journal of Operational Research
An Information-Collecting Drone Management Problem for Wildfire MitigationarXiv preprint

MIT Lincoln Laboratory

Prior to Princeton, I contributed to applied machine learning research at MIT Lincoln Laboratory.

Learning-based long-range axon tracing in dense scenesISBI 2018. CNNs for tracing axons through neurons.
An open-source noise dosimeter for evaluating auditory environmentsJASA