Skip to content

Projects

Things I built from problems I noticed.

Most of these came out of something that annoyed me. Filter by type of work or by domain.

Type
Domain

10 projects

PersonalComing soon

BoxIt

College storage is expensive, inconvenient, and low-trust. Students with spare space would rather earn money hosting it.

Systems/BackendFull-StackInfra/Cloud
Personal

Liberation Day and After

I wanted to see how one big macro shock ripples through sectors and reshapes market structure.

QuantDataApplied-Research
Personal

EquiTable

Atlanta's food-insecurity problem was driven by disconnected food banks and missing information. I wanted to bridge that gap.

ML/AIFull-StackInfra/Cloud
Hackathon

FirstWave

Emergency demand is predictable, like Friday nights in the Bronx or summer weekends in Brooklyn. Ambulances still wait for the call instead of moving ahead of it.

Full-StackFrontendDatadraft
Personal

Benchwarmer

I follow the NBA and I like prediction. I wanted to treat it like a market: a backtested engine against live prices, not gut calls.

QuantML/AIData
Research

Steering Truth in LLMs

The standard ways to find a 'truth direction' in a model, like linear probes and contrastive mean differences, leave signal on the table. I tried to feature-engineer better directions and steer on them.

ML/AIApplied-Researchdraft
Personal

jobmaxxing

Internship postings are scattered across a dozen boards and go stale fast. I was tired of refreshing all of them, so I built one feed that watches them for me.

Systems/BackendDataInfra/Clouddraft
Personal

IMC Prosperity 4

I wanted to see whether my read on a market would hold up against a live order book, not just a clean backtest.

QuantDatadraft
Hackathon

clearRx

Whether two prescriptions interact shouldn't come down to a provider's memory, and a lookup tool only helps if you can trust its answer.

ML/AIFull-Stackdraft
Research

Aquatic Sustainability

Water problems go undetected in the places with the least monitoring data. I wanted to see watershed risk and access gaps on a map before they turn into emergencies.

ML/AIDataApplied-Researchdraft