Stefen Ewers
ML engineer. Came up through fintech and real estate — now building AI that works in both.
YardToonz views, week one
Zero paid promotion.
Airbnb model accuracy
Live API, sub-200ms.
Reporting time cut at Divvy Homes
Automated AWS pipeline.
Industries with production ML
Fintech, real estate, blockchain.
Finance. Real estate. Now ML.
Experience across real estate, fintech, and proptech — building predictive financial models at Divvy Homes, running fraud detection systems at Wells Fargo. Analytical work grounded in real systems, not theory.
Now completing an M.S. in Computational Analytics (ML & AI) at Georgia Tech, while interning at Knovel Protocol building NLP pipelines on blockchain data. Originally from Jamaica. Based in Atlanta.
Ewers Scholar Initiative
Donating laptops to top students passing from St. Richards Primary to Campion College.
Campion is the best secondary school in Jamaica. These students earned their place — the initiative makes sure technology isn't the barrier to what comes next.
Domain expertise is part of the model
I spent years in fintech and real estate before applying ML to either. That context shapes which features matter, which outputs are actionable, and where a model can actually fail.
Specificity is a performance requirement
The Jamaican Airbnb model outperforms a generic one because it was built for that market. YardToonz works because Jamaican content has its own aesthetic. Context isn't decoration — it drives accuracy.
Ship to learn, not to finish
A deployed API returns signal a notebook never will. The underwriting engine isn't done — it's running. I build toward real feedback, not a final spec.
Design for the decision, not the metric
R²=0.84 only matters if it helps someone price a rental. I optimize for what the person on the other end actually does with the output.
Let's Build
If you're working on something interesting — or need someone who can build and think — reach out.
This site is partially powered by AI systems I built.