Portrait of Alexander Hewitt

Alexander Hewitt

Data scientist & operations research analyst

From the flight line to the forecast.

I turn operational data into models people can use. My path runs through aircraft maintenance, financial decisions, federal forecasting, and products I built from scratch.

A conceptual 3D model connects an aircraft, a forecast curve, and a software interface.
Operations. Models. Products.Concept illustration

Now at USCIS Operations Research Analyst, GS-14

Edmond, Oklahoma Remote-first

2011 to today

Different systems.
The same curiosity.

Choose a chapter to see the problem, the method, and the result. The setting changes. The question stays: how can this work better?

Chapter 7 of 8

July 2011 – January 2014

Start where the work happens.

U.S. Air Force, Tinker AFB

Data Analyst / Maintenance Scheduler, GS-11

I built performance reports and improved the data behind maintenance scheduling. The first lesson: useful analysis starts with reliable operational records.

3,000+part numbers in a new labeling system

Supported 10+ process improvements across maintenance operations.

  • Operational reporting
  • Process improvement
  • Labor-hour analysis
Making the system visibleTechnical illustration

A consistent part label connects the physical item to its operational record.

January 2014 – April 2017

Put a number on the trade-off.

Self-employed, Tulsa

Real Estate Investor / Analyst

I evaluated acquisition, renovation, and resale decisions. Cost-benefit analysis became a practical tool for choosing properties, contractors, and project scope.

50+properties evaluated

Managed 15+ projects and reduced contractor costs by 20% through vendor selection and planning.

  • Cost-benefit analysis
  • Vendor comparison
  • Project planning
Underneath the purchase priceTechnical illustration

Acquisition cost, renovation scope, and resale assumptions define the decision.

April 2017 – March 2020

Think in probabilities.

HFLLC, Tulsa

Day Trader / Portfolio Analyst

I analyzed derivatives activity and built rule-based, time-series decision models. The work required explicit assumptions, disciplined risk analysis, and decisions under uncertainty.

$15M+derivatives activity in a calendar year

A foundation in probabilistic reasoning that later carried into forecasting and scenario analysis.

  • Time-series analysis
  • Risk analysis
  • Decision rules
One decision, many outcomesIllustrative distribution

Time-series signals inform a decision. They do not guarantee an outcome.

March 2020 – August 2020

Make the numbers reconcile.

Bank of Oklahoma, Tulsa

Security Services Operations Specialist I

I processed transactions and reconciled client funds. Accuracy, controls, and traceable reporting mattered at every handoff.

$30M+daily client funds processed and reconciled

Supported audit, compliance, and operations through reports and issue resolution.

  • Reconciliation
  • Data controls
  • Issue resolution
A traceable path through the dataProcess model

Transaction records enter a controlled process for accurate reconciliation.

August 2020 – August 2021

Find the constraint in the flow.

U.S. Air Force, Tinker AFB

B-52 Aircraft Maintenance Data Analyst, GS-11

I returned to aircraft sustainment with a broader analytical toolkit. I examined maintenance flow, workforce performance, and the causes behind production bottlenecks.

B-52maintenance operations and production analysis

Analyzed avionics, sheet metal, and electrics. Improved recurring reports with Excel macros.

  • Root cause analysis
  • Excel / VBA
  • Production metrics
A system of connected workAircraft schematic

Maintenance records connect shop activity, workforce capacity, and production flow.

August 2021 – August 2022

Turn recurring defects into a signal.

U.S. Air Force, 565th AMXS

B-52 Aircraft Lead Data / Management Analyst, GS-12

I designed a Functional Check Flight defect-tracking program. Connecting events across parts exposed recurring patterns and supported changes to maintenance flow.

700+Functional Check Flight events tracked

Identified recurring patterns in 10+ parts. Reduced post-dock flow by 5+ days, approximately 10%.

Also identified 35,000 production hours at risk from proposed policy changes.

  • Reliability signals
  • Capacity analysis
  • Leadership briefings
From events to a reliability signalProcess model

The tracking program connected 700+ check-flight events to recurring part failures.

August 2022 – present Current role

Forecast at federal scale.

U.S. Citizenship and Immigration Services

Operations Research Analyst, GS-14

I lead revenue and cost forecasting across 30+ revenue-generating forms. I build the data pipelines, scenario models, and reports behind executive decisions.

$7B+annual revenue and cost forecasting supported

Automated recurring reporting, saving approximately 24 workdays per year.

  • Python
  • SQL
  • Databricks / PySpark
  • SAS / Tableau
Explore a forecastIllustrative model

Move the slider to change an illustrative growth assumption. The curves show a scenario and uncertainty, not USCIS data.

March 2026 Education

Test the model. Question the score.

Eastern University

MS in Data Science, 4.0 GPA

My capstone joined 2024 flight records with historical weather and compared six regression models. I removed post-departure leakage and evaluated pre-departure signals only.

5.85Mflight records in the capstone pipeline

LightGBM test MAE: 21.7 minutes. Median absolute error: 12.6 minutes. R²: 0.077.

  • Python
  • LightGBM / XGBoost
  • scikit-learn
  • Streamlit
Compare the test resultsRecorded evaluation

Minutes of error · lower is better · bars start at zero

These are the corrected test results. Modest predictive performance is part of the result, not a number to hide.

Independent products

Models that became tools.

I also build products from the first data question to the interface people use. These are two examples you can explore today.

Zollera's live homepage compares a contractor quote with a local price range.Live product preview

Document intelligence

Zollera

A contractor quote, made easier to question.

I built a platform that turns PDFs, photos, and scans into comparable line items. A tiered LLM pipeline handles extraction and provider failover.

PII detection, redaction, normalization, and review controls surround the model so the output can support a useful decision.

  • Next.js / TypeScript
  • PostgreSQL
  • Multi-provider LLMs
Explore Zollera
HighThree's live homepage presents an illustrative retirement income breakdown.Live product preview · example scenario

Financial modeling

HighThree

Make the long-term trade-off visible.

I built federal retirement tools that project salary, service, and pension income. Present-value models and scenario comparisons help users examine career decisions.

All calculations run in the browser. Salary and service inputs stay on the user’s device.

  • JavaScript
  • Scenario analysis
  • Present-value models
Explore HighThree

Flight delay prediction

71 pre-departure features. Six evaluated models. A Streamlit interface with per-prediction feature attribution.

Explore the evaluation

The thread through the work

Understand the operation.
Then build the model.

I’m most useful where the data is messy, the decision matters, and someone needs to own the path from raw records to a usable result.

I’m interested in senior data science, operations research, and analytics engineering roles focused on forecasting, simulation, or predictive maintenance.

Start a conversation

Education

MS, Data Science
Eastern University, 2026 · GPA 4.0

BBA, Supply Chain Management
University of Oklahoma, 2011

Methods & tools

Forecasting, Monte Carlo simulation, applied statistics, feature engineering, and scenario analysis.

Python, SQL, R, SAS, Databricks, PySpark, Tableau, Power BI, and Excel / VBA.

Professional context

Lean Six Sigma Green Belt. U.S. citizen. Tier 5 background investigation favorably adjudicated in 2025. No active security clearance.