Spain Powell

Data science / systems / product

Models that move the work.

I build predictive models, analytics systems, and practical products that turn complex data into decisions people can use.

View selected work Start a conversation
Current focus Active
Applied analytics → AI systems From operational models to reliable agent workflows.
33.66°N 117.91°W Costa Mesa, CA
Scroll to trace the work 01 / 04
Under 5 min Model run time per location
$1.5M / mo Reported operational savings
$24M / yr Projected pickup opportunity
3+ years Building impact-oriented systems

01 / Selected work

Evidence over ornament.

Different surfaces, same instinct: structure the mess, make the decision legible, and leave behind something that can run.

01 Logistics analytics / FedEx

Courier productivity model

Forecasting for time-commit-centric logistics operations.

Extended a Python model to forecast new facility performance up to 12 months ahead. Replaced a manual engineering workflow that took hours per station with a model run under five minutes per location.

PythonSQLForecastingOperations
Reported impact $1.5M/mo $16M monthly upside identified
02 Building / Independent

Win The Numbers

A fantasy football companion built around roster-aware value.

Mobile product that combines dynasty player values, a trade calculator, roster-aware trade ideas, and team analysis around each league's settings and direction. Built with a monetization and launch system around the product.

PythonReact NativeSports dataAdMob
Visit product site
Signal Product Data, algorithms, UX, and launch path
03 Shipped / Clemson

Baseball analytics

From real-game tracking data to coaching decisions.

Built a Streamlit application for Clemson Olympic Sports, translating real-game sports tracking data into performance metrics, scouting reports, and post-game breakdowns for coaching decisions.

StreamlitSports dataPythonAnalytics
Audience Coaches On-demand, decision-ready reporting
04 Active / TheBrain

Claude / Codex harness

An operating system for reliable agent-assisted work.

Designed shared doctrine, runtime mappings, source boundaries, hooks, evals, runbooks, manifests, and review loops that make AI-agent workflows more repeatable and more accountable.

Agent systemsEval designProvenanceRunbooks
Design principle Verify Promote patterns only after evidence

Also in the lab

FoamFinger + Braisen

Two structured product laboratories exploring sports pick'em, social cooking, monetization, compliance-aware planning, and phase-gated execution.

02 / Method

Make the complex usable.

The useful output is rarely the model alone. It is the model, the interface around it, the evidence behind it, and the operating path that keeps it useful.

  1. 01
    Frame the decision

    Start with the operator, constraint, and outcome. Make the real question explicit before selecting a tool.

  2. 02
    Structure the signal

    Pull together messy data, source context, and stakeholder knowledge into something that can be inspected.

  3. 03
    Ship the surface

    Turn analysis into a model, dashboard, report, or product that fits the rhythm of the person using it.

  4. 04
    Leave a verifier

    Document the assumptions, add the check, and make the next run easier to trust than the last.

03 / About

A builder between the model and the moment it matters.

I’m Spain, a data scientist and engineer based in Costa Mesa, California. My work sits between predictive modeling, analytics engineering, operations, and product-minded execution.

I like the point where a messy system becomes legible: a forecast a team can act on, a dashboard that changes the conversation, or a workflow that makes an AI agent more reliable.

Toolbox

Python · SQL · R · React Native · Streamlit · Power BI · Tableau · Excel · AWS · GitHub

Education

M.S. Data Science & Analytics
Clemson University · 2025

B.S. Computer Science
University of California, Irvine · 2022

Currently looking toward

Data science and AI roles where predictive models, analytics systems, and data products create measurable real-world impact.

04 / Contact

Have a hard problem that needs a useful next step?

I’m interested in teams building data products, applied AI, and systems that hold up beyond the demo.

Connect on LinkedIn