Stefen Ewers

Software engineer building reliable web applications, APIs, and data systems.

I work primarily in TypeScript and Python, with React and Next.js on the frontend and APIs and data pipelines behind it. My analytics background shapes how I model complex domains, validate boundaries, and test critical logic.

View engineering projectsGitHubRésumé
now

Software Engineer · Sep 2024–Present

Knovel Protocol · previously Data Analyst, May–Sep 2024

before that

Divvy Homes · iHeartMedia

Predictive models, data pipelines, SQL

alongside

M.S. Analytics, Computational track

Georgia Institute of Technology · 2025–2028

Ask Winston

He answers in a sentence or two. Turn on the speaker for audio.

// selected engineering work

Testero

Deployed

Domain modelling and correctness

An origination system for home equity investment. It keeps one canonical case state, traces every material fact back to the document, page and block it came from, and derives readiness with deterministic rules.

  • Framework-independent domain core
  • 40 rule and invariant tests
  • Contradiction-aware evidence

TypeScript · Next.js · React · Zustand · Zod · Vitest

Live demoSource

Independent project on synthetic data. No production deployment.

Testero case view for a synthetic file: a readiness panel reporting Not ready with two blockers, a decision state column listing blockers and a policy conflict by rule ID, a ranked next-best-actions list showing which rule each action unblocks, and an evidence panel flagging a contradiction in ownership vesting between the application and the recorded deed.
Meridian experiment workspace: baseline and proposed fleet configurations side by side, the assumptions in force, and a control rail for fleet size, charger count, airport battery reserve and dispatch policy.

Meridian

Complete

Backend services and simulation

  • Python FastAPI service, TypeScript frontend
  • 25 Pytest tests
  • Cross-process reproducibility

Python · FastAPI · SimPy · OR-Tools · +4

Case studySource
Trendr dashboard: strategy versus buy-and-hold equity curves over the held-out test period, the signal distribution across long, short and flat, and a table of the most recent daily signals.

Trendr

Deployed

Data and ML pipelines

  • Look-ahead leakage audits
  • 56 Pytest tests
  • Walk-forward, cost-aware backtesting

Python · pandas · NumPy · scikit-learn · +5

Live demoCase studySource
ML Internals running K-means: a canvas of points assigned to three coloured centroids, a terminal log reporting convergence, playback controls with speed and a k slider, and an insight panel describing the current step.

ML Internals

Deployed

Frontend and interaction

  • Five algorithms implemented directly
  • Per-algorithm state machines
  • Canvas rendering, dynamic imports

TypeScript · React · Next.js · Canvas

Live demoCase studySource
All projects, including Mandate, Current and Winston
// experience

Knovel Protocol

  • Software Engineer

    Sep 2024 – Present

  • Data Analyst

    May 2024 – Sep 2024

San Francisco Bay Area · Remote

Builds recommendation and evaluation features for a blockchain data platform.

  • Built the similarity and recommendation layer behind content discovery and search, which decides what a user gets surfaced.
  • Defined the business logic and evaluation criteria applied to anomaly-detection output: what counts as irregular in blockchain transaction data, and which flags are worth an operator’s attention.
  • Built the post-release evaluation process the team measures system output against, which is what ranks the next changes and what settles whether a data-enabled feature beat a simpler rule.

Recommendations · Anomaly detection · Evaluation

Divvy Homes

  • Associate

    Jan 2023 – Apr 2024

  • Analyst

    Apr 2022 – Jan 2023

San Francisco Bay Area · Remote

Predictive models and data pipelines behind acquisition and portfolio decisions at a real estate technology company.

  • Predictive financial models behind two live decisions: which homes to buy, and how the existing 1,000+ property portfolio was performing.
  • Rebuilt recurring reporting as an automated pipeline on AWS, cutting reporting time by 40%. The first time the answer to a manual process was to write software.
  • Sat between operations, finance and engineering, which is where the requirements for any of it came from.

Python · SQL · AWS pipelines · Financial modeling

iHeartMedia

  • Data Analyst II

    Feb 2021 – Apr 2022

  • Data Analyst, Media & Audience Insights

    Jan 2020 – Feb 2021

Tampa / St. Petersburg, Florida · On-site

Audience and campaign data in SQL, and the reporting the planning teams worked from.

  • Audience and campaign analysis in SQL for the teams planning media spend, and the recurring reporting behind it.
  • Two years and a promotion in the same problem space. Where the habit started: a number is worth what the decision it changes is worth.

SQL · Audience analytics · Reporting

Technical toolkit
Languages

TypeScript · JavaScript · Python · SQL

Frontend

React · Next.js · HTML / CSS · Tailwind CSS · Canvas

Backend & APIs

Next.js API routes · FastAPI · REST APIs · Zod · Pydantic

Data & ML

pandas · NumPy · scikit-learn · LightGBM · SHAP · SimPy · OR-Tools

These libraries are from the projects below. Professionally, the related work is the recommendation layer at Knovel Protocol and the evaluation criteria applied to anomaly-detection output.

Testing & tooling

Vitest · Pytest · node:test · Playwright · Git / GitHub · Vercel · AWS

Vitest, Pytest, node:test and Playwright are used in the projects below. AWS is from the reporting pipeline at Divvy Homes.

Technical foundation
Aug 2025 – Spring 2028

Georgia Institute of Technology

M.S. Analytics, Computational Analytics track

Statistical modeling · Optimization · Machine learning · Deep learning

Jan 2025 – Dec 2025

Massachusetts Institute of Technology

Graduate Certificate, Data Science and Machine Learning

Causal inference · Deep learning · Recommender systems

2019

University of South Florida

B.S. Finance / Business Analytics

// about
Stefen Ewers

From analytics to software engineering.

At Knovel Protocol I build recommendation and evaluation features for a blockchain data platform. Before engineering, I worked in analytics at iHeartMedia and Divvy Homes, where rebuilding a recurring AWS reporting workflow moved my work from analysis toward software.

That background is why I start with the domain model and the tests. The problems I like are the ones where the rules are genuinely complicated, the data disagrees with itself, and a system that is quietly wrong costs more than one that fails loudly.

Originally from Jamaica.

Outside work

The Ewers Scholar Initiative

Laptops for the top students passing from St. Richards Primary to Campion College, the best secondary school in Jamaica. They earned their place. The laptop is so a computer isn't the thing standing between them and it.

A $1,000 grant also goes to a Jamaican student at the University of South Florida each year.

Azaylia Palmer receiving her Lenovo laptop from Stefen Ewers at St. Richards Primary
Azaylia Palmer with her teacher holding her new Lenovo laptop
// contact

Let's talk

I'm interested in software-engineering work on full-stack applications, APIs, data-intensive systems, or applied ML. If you're building something technically demanding and useful, get in touch.

stefenewers@gmail.comGitHubLinkedIn