Engineering ML02
Model-builder workflows
Surrogate models, design-of-experiments tooling, and optimization loops that make expensive engineering analyses easier to explore.
PythonDOEsurrogates
Applied research · scientific ML · engineering systems
I turn messy engineering data into models people can use.
I’m Isaac Rose, a simulation and modeling engineer who builds the bridge from physical intuition to reliable software and data-driven decisions.
data
model
decision
IRbuild · test · learn
01Physics-informedmodels grounded in how systems behave
02Data-drivenfeatures, uncertainty, and evaluation that hold up
03Deployedusable tools instead of notebooks that stop at a plot
Selected work
Research instincts, engineering follow-through.
Projects where the interesting part is not just the model—it’s the loop around it: data quality, validation, explanation, and a user who needs a useful answer.
Featured · applied ML↗
Flight Reliability
A risk-aware flight decision engine that turns operational history, weather, and itinerary context into explainable disruption signals.
FastAPINext.jsforecastingdecision support
Physics & validation03
Damper modeling
Measured behavior, physical models, and validation workflows connected into engineering tools that support better decisions.
simulationvalidationvehicle dynamics
The working loop
Useful models are a team sport.
01
Frame the question
What decision should this model improve? What does “wrong” cost?
02
Respect the data
Build the pipeline, find the leakage, understand the missingness, keep the baseline.
03
Measure uncertainty
Calibrate the output, expose the drivers, and make confidence part of the interface.
04
Ship the learning
Put the result where someone can use it, then let real feedback shape the next experiment.
Let’s build something that learns
Have a hard system, a noisy dataset, or both?
I’m interested in applied research where physical reasoning and machine learning make each other better.