What I do
I build forecasting, simulation, and predictive models on large operational datasets — and I ship the pipelines and dashboards that put them in front of the people making decisions. Most of my work has been in environments where the data is messy, the stakes are real, and the model has to be explainable to a non-technical audience.
Ten-plus years across federal service (GS-14 Operations Research Analyst at USCIS), Air Force depot maintenance analytics at Tinker AFB, banking, and derivatives trading. MS in Data Science, 4.0 GPA. Lean Six Sigma Green Belt.
Selected work
Zollera — contractor quote analysis platform
A production web application that ingests contractor quotes as PDFs, photos, or scans and extracts them into structured, comparable line items, so a homeowner can see where a quote sits in the market. I designed and built it end to end: 83 API routes and 61 pages on Next.js 15 with TypeScript and PostgreSQL, including authentication, role-based admin tooling, rate limiting, and audit logging.
The extraction core is a tiered, multi-provider LLM pipeline with automatic failover — Gemini for native PDF handling, then vision models — running under one shared time budget so a slow tier can't starve the fallback. Around it sits PII detection and redaction before any model call, price outlier and fraud-signal checks, duplicate-document detection, unit and financial normalization, and an admin QC workflow for bulk uploads.
HighThree — federal retirement analytics
A retirement analytics suite for federal employees under FERS: 43 reference pages and 59 locality-pay pages. The core model projects salary and service year by year, then discounts the resulting annuity stream to present value. Companion tools cover break-even age, claiming timing, survivor and COLA effects, TSP drawdown, and side-by-side scenario comparison.
Deliberately built with vanilla HTML, CSS, and JavaScript — no framework, no build step, no server. Every calculation runs in the browser, so a user's salary and service history never leave their device. Hardened with a strict Content-Security Policy and built for search discovery with programmatic locality-pay pages and generated social cards.
Flight departure delay prediction
An end-to-end regression pipeline over 5,854,077 U.S. domestic flights from 2024, joined to historical weather observations at the top 100 airports. Six models evaluated — linear, regularized, random forest, XGBoost, and LightGBM — with LightGBM deployed. 71 pre-departure features spanning schedule, route history, and weather. Deployed as an interactive Streamlit app with per-prediction feature attribution.
Reported honestly: test MAE 21.7 min, median absolute error
12.6 min, R² 0.077. An earlier iteration scored much higher only because it
leaked post-departure fields such as carrier_delay — values
recorded after a flight is already late. I found the leakage, removed it,
and re-evaluated on pre-departure signals only. Predicting delays before
departure is genuinely hard; the median error is the figure I'd stand
behind, and the notebook documenting the leak is in the repo history.
Background
USCIS — Operations Research Analyst (GS-14), 2022–present. Lead revenue and cost forecasting across 30+ revenue-generating forms supporting analysis of more than $7B annually, using Python, SQL, Databricks, and PySpark. Built and maintain the forecasting models, scenario analyses, and automated reporting pipelines behind executive decision support.
U.S. Air Force, Tinker AFB — B-52 maintenance analytics, 2011–2014 and 2020–2022. Designed a Functional Check Flight defect-tracking program covering 700+ events, identifying recurring failure patterns across 10+ parts and cutting post-dock flow days by roughly 10%. Quantified the labor-capacity impact of proposed policy changes, flagging 35,000 production hours at risk.
Education. MS in Data Science, Eastern University (2026), 4.0 GPA. BBA in Supply Chain Management, University of Oklahoma. Lean Six Sigma Green Belt.
What I'm looking for
Senior data science, operations research, or analytics engineering roles — remote-first, working on forecasting, simulation, or predictive maintenance problems at meaningful scale. I'm most useful where there's real operational data, a decision that depends on it, and room to own the pipeline from raw ingestion through to the model leadership trusts.
I'm a U.S. citizen with a Tier 5 (Top Secret/SCI-equivalent) background investigation favorably adjudicated in 2025. I do not hold a current active clearance — no need-to-know basis at present — and am eligible for sponsorship or reinstatement. Previously served on-site in a cleared Department of the Air Force depot environment.