Experience

10+ years building production statistical and machine learning systems.

Machine learning engineer with 10+ years of experience delivering production statistical systems across media and advertising measurement, defense, and consulting. My focus is the methodology behind population-level estimates and the verification machinery that lets those systems change safely.

Roles

Senior Machine Learning Engineer

2025 – present

VideoAmp · Remote

Own the production pipelines that calibrate audience-measurement panels to the population: population targets, household weighting, and the validation that gates every release. Engineering owner for the weighting workstream of a cross-team methodology refresh, working with methodology, data science, and engineering partners.

• Replaced a legacy weighting system by first reproducing it. Rebuilt the household-weighting pipeline config-first and reproduced a published week of the legacy system’s output before moving it onto a new raking solver. Reconciliation gaps were traced to root causes in the legacy code instead of being written off as solver noise.
• Measured what a method change does. Built behavioral-preservation reports that test whether new weights shift what weighting should leave alone, such as program rankings and viewing-time distributions. They became the team’s baseline for judging weighting changes.
• Migrated the team’s core statistical pipelines into a shared production monorepo: Bayesian household-composition inference, population targets from census and survey data, calibrated household weights, and new streaming-TV panels. Each followed the same scaffold → validate → promote path, with availability checks, schema enforcement, input snapshots, monitors, and parallelization that substantially shortened end-to-end runtime.
• Methods. Raking (iterative proportional fitting) and calibration weighting formulated as convex optimization; Bayesian inference over household composition; reproducible probabilistic assignment with deterministic hashing; gradient-boosted models with dated versioning and freshness guards that block inference when inputs post-date the training data.
• Verification layer. Validation gates that fail on distributional drift from historical baselines, a regression harness that re-solves frozen fixtures through the production code path at tight numerical tolerance, all-or-nothing writes for large parallel fan-outs, dry-run-by-default delivery runners, and run snapshots for reproducible replay.
• Cross-team leadership. Put design options and trade-offs, such as solver fit versus weight dispersion, to methodology and engineering stakeholders for a decision before implementation. Chair a weekly sync between the population-targets and weighting workstreams, onboard engineers through code review, and wrote a glossary that translates methodology into engineering terms.
• AI-assisted engineering. Built an agent skill, with its own evaluation set, that the team uses to file tickets from notes. Run coding agents under a written playbook of verification rules, and review pull requests with parallel reviewer agents whose comments are fact-checked against the code before they are posted.

Snowflake · Snowpark · Airflow · Python · cvxpy · scikit-learn · XGBoost · Datadog

Lead / Senior Machine Learning Engineer

2021 – 2024

Tiger Analytics · Remote

Client-facing ML engineering: model training, evaluation, deployment, monitoring, and explainability.

• Designed a standardized ML pipeline template, with preflight data checks, data-quality checks, training, promotion, and monitoring as ordered stages, adopted across engagements.
• Led migration of production models from SAS to Python, validating each against its baseline with row-level output comparison before cutover.
• Built a retrieval-augmented question-answering assistant on an LLM and a vector store.

Databricks · SageMaker · MLflow · PySpark · XGBoost · LightGBM

Senior Machine Learning Engineer

2020 – 2021

Kratos Defense & Security Solutions · Colorado Springs, CO

• Built a convolutional neural network that combined multiple telemetry sources to identify environmental causes of performance degradation, shortening time to diagnosis.
• Designed cloud data-ingestion pipelines feeding automated analysis.

PyTorch · AWS · Python

Senior Machine Learning Engineer

2018 – 2020

DISH Network · Denver, CO

• Built a self-serve impressions forecasting tool that replaced manual estimates: users select any combination of segment attributes and a horizon and receive a forecast.
• Modeled each selection as a time series with trend and seasonality components and monitored forecast accuracy with MAPE.
• Mentored junior staff in error analysis, machine learning, and cloud architecture.

Python · AWS · Databricks · Docker

Data Scientist / Machine Learning Engineer

2017 – 2018

Kratos Defense & Security Solutions · Colorado Springs, CO

• Built a time-series anomaly-detection model with automated alerting that substantially reduced detection time.
• Developed classical and deep-learning models for pattern recognition in signal data, and migrated analytics storage to a cloud data lake.

Python · AWS · scikit-learn

Additional experience

• Assistant Professor, Machine Learning, Colorado Technical University (remote)
• Senior Field Applications Engineer, Fortemedia (Sunnyvale, CA)
• Field Applications Engineer, Microchip Technology (San Jose, CA)
• Field Applications Engineer / Field Support Engineer, SRI International (Menlo Park, CA)

Education

• Doctor of Computer Science, Big Data Analytics, Colorado Technical University
• M.S., Data Science, Southern Methodist University
• B.S., Engineering, Harvey Mudd College

Certifications

• Databricks Certified Machine Learning Professional (2024)
• AWS Certified AI Practitioner (2024)
• AWS Certified Machine Learning – Specialty (2022–2025)

Skills

Methods. Survey weighting and raking (IPF) · calibration estimators · Bayesian inference · convex optimization · population estimation · time-series forecasting · gradient boosting · deep learning for time series

Verification & reliability. Distributional validation and drift detection · regression testing for nondeterministic solvers · reproducibility and run versioning · model monitoring

AI-assisted engineering. Coding agents run under written verification playbooks · evaluated agent skills · multi-agent code review

Languages & libraries. Python · SQL · PySpark · PyTorch · scikit-learn · XGBoost · LightGBM · cvxpy · statsmodels · NumPy · SciPy · pandas

Platforms. Snowflake (Snowpark, Container Services, Model Registry) · AWS (SageMaker, Lambda, Athena, EMR, EKS, S3) · Databricks · Airflow · Docker · Kubernetes · GitHub Actions · GitLab CI · MLflow · Datadog