Candidate for Machine Learning Research Engineer
I build models that help uncertain systems become useful.
My primary fit is applied research: building models, evaluating them honestly, and turning uncertain predictions into useful decisions.
Research
that survives
contact with reality.
Why WindBorne
The work sits at the intersection I keep coming back to.
I like problems where the data is incomplete, the physics still matters, and a prediction only earns its keep if it improves a real decision.
That is what draws me to WindBorne. Weather is a rich, consequential modeling problem: spatially distributed observations, changing regimes, imperfect sensors, and a long tail of failure modes. The challenge is not simply to fit a model; it is to build a system that knows what it knows, communicates uncertainty, and improves as new evidence arrives.
I bring a simulation engineer’s instinct for mechanisms and a builder’s instinct for closing the loop. I want to work close to the data and the models, with enough product awareness to make research useful outside a notebook.
Project evidence / 01
Flight Reliability
A full-stack experiment in turning heterogeneous operational and environmental data into calibrated, explainable risk signals.
Which itinerary is actually the safer choice?
Instead of optimizing for fare and departure time alone, the product estimates delay, cancellation, connection, and recovery risk.
- Data layerhistorical operations, route context, airport behavior, weather features
- Model layerbaseline/statistical scoring with a staged path to calibrated ML
- Product layerranked comparisons, explanations, recovery flows, source-backed guidance
- DeploymentNext.js · FastAPI · PostgreSQL · Redis · Docker · Caddy
The project is deliberately honest about the boundary between shipped behavior and future research. That discipline—baselines, calibration, explanations, and deployment—is the part I would carry into atmospheric prediction.
Next page
Let’s make the forecast useful.
The application package is grounded in the work shown here. I’m happy to send the current résumé and written responses directly while I keep refining the public case study.