I build useful experiments at the intersection of machine learning, full-stack engineering, and how people think. Lately: graphs, language models, and the psychology of information.
A few experiments that started as a quiet 'I wonder if…' and turned into real tools. Open any card for the technical story.
Less resume, more desk. Poke around — every object is a small, honest clue about how I actually work.
Hover or tap the objects on the desk. A few of them have opinions.
Internships and campus work where the common thread was the same: take unreliable, real-world data and make it trustworthy, observable and useful.
Data & ML Engineering Intern
Made a fashion catalog trustworthy. Built data-validation pipelines and ML normalization so millions of garment records — and the images attached to them — stayed clean, consistent and monitorable.
Full-Stack & Data Intern
Turned operational data into decisions. Built React/Node dashboards over ERP and WMS data, ran SQL audits to catch what spreadsheets missed, and automated quality control with OpenCV.
Campus Engineering & Infrastructure
Built tools the whole campus uses. Shipped Django/React/PostgreSQL products, instrumented them with Mixpanel, set up CI/CD, and helped plan real infrastructure — including a campus-wide Wi-Fi heat-map initiative.
Alongside engineering, I run an ongoing research thread where the questions are human — and the methods are a mix of psychology and machine learning.
An ongoing project at the intersection of psychology and machine learning — studying not just what misinformation is, but how people read it, trust it, and get moved by it. It pairs survey design and analysis with an ML model that detects manipulation mechanisms in text.
How perception, trust and manipulation actually operate on a reader.
Designing and analyzing studies of how people judge information.
A multi-label BERT classifier for manipulation mechanisms in text.
Give me an ambiguous question and a pile of messy data and I'm happy. Most of what I build follows the same arc — take something unstructured, understand the data and the people behind it, and turn it into something useful, interactive, or just fun to poke at. A psychology minor quietly shapes all of it: I care as much about how people read an interface as how the model underneath it works.
The best problems show up as ambiguity and messy data. I like the part where you figure out what the question even is.
A model or a query is only useful if someone can actually explore it. I care about the interface as much as the pipeline.
Attention, trust, how people actually decide — a psychology minor rewired how I think about the humans on the other side.
I'd rather build a useful experiment than write a long plan. Most of these projects started as a quiet 'I wonder if…'.
A working set, not a buzzword wall — these are the tools I've actually shipped with across frontend, backend, data and ML.