Software engineer (MS, AI specialization), building full time after ten years running web platforms and analytics for a national youth soccer organization.
Across more than ten years in the youth soccer ecosystem, my work has included managing websites and web platforms serving 1M+ annual viewers, writing website code, maintaining site architecture and performance, building analytics and executive reports, and translating business requirements into technical tasks.
I pair that domain experience with an MS and BS in Software Engineering and hands-on portfolio work in Go, TypeScript, Python, SQL, PostgreSQL, APIs, data pipelines, Docker, testing, CI/CD, and AWS-oriented architecture. These repositories demonstrate the engineering work behind my career transition; they are independent portfolio/reference implementations, not systems built for any employer.
| Reviewer path | First project | What it demonstrates |
|---|---|---|
| Software engineering | Learning Center Reference · static tour | Working Go/Next.js/PostgreSQL learner and administrator workflow with OIDC verification, RBAC, persisted progress, derived eligibility, tests, and accessibility automation |
| Data engineering | Fan Unification Platform · live dashboards | Prefect-orchestrated ingestion, explainable identity resolution, SCD2, dimensional modeling, SQL quality gates, measured linkage, and BI-ready marts |
| Machine learning evaluation | Deep Learning Portfolio Lab · evidence card | Frozen-split recommender evaluation with preregistered hypotheses, leakage controls, slice tables, and CI-regenerated evidence, including two published negative results |
My current technical focus is deliberately narrow: multi-role learning and eligibility workflows, and trustworthy identity/data integration across disconnected soccer systems.
A fictional soccer-federation learning and eligibility platform. Its working vertical slice verifies identities, resolves database roles, persists course enrollment and ordered lesson progress, projects a learner dashboard, and derives an administrator compliance view from expiring credentials and holds. The stack is Go, Next.js/TypeScript, PostgreSQL, OpenAPI, Docker, and GitHub Actions.
A Prefect-orchestrated identity-resolution and dimensional-warehouse pipeline over four synthetic source systems, including a Salesforce-shaped API. It demonstrates incremental and idempotent ingestion, quarantine, explainable matching, SCD2, SQL quality gates, Redshift-oriented design, scheduled CI, and BI-ready marts. Linkage accuracy is measured against generated ground truth, including a published result where the first probabilistic approach lost to the deterministic baseline.
An auditable FastAPI reference system for public sanctions-list ingestion, explainable entity matching, information extraction, retrieval, evaluation, and human-review exports. It fails closed when verified source data is unavailable and clearly separates fictional examples from public-source adapters.
One deep-learning experiment carried all the way through and fourteen reviewed contracts held as backlog. The temporal Steam recommender compares popularity baselines, BPR matrix factorisation, and a two-tower retriever on a frozen 2022 split with per-user event-time leakage controls, hypotheses preregistered before training, three seeds, slice tables, and a CPU retrieval service. The two-tower model beat matrix factorisation and reached never-seen releases; a recent-popularity baseline won the sale-season test window, and a preregistered hybrid follow-up failed too. Both negative results are published, and the report, evaluation card, and recruiter FAQ figures are regenerated from retained run manifests in CI. Python, PyTorch, DuckDB, PyArrow, FastAPI, strict mypy, GitHub Actions.
Four event-driven Python services use Kafka, PostgreSQL, Redis, OpenAPI, WebSockets, observability, Docker, and Kubernetes to turn synthetic surveillance reports into tracks and advisory alerts. The repository includes contract, integration, end-to-end, performance, and failure-mode evidence, including a model improvement that did not replicate under shifted scenarios.
| Project | Status and engineering signal | Implemented stack |
|---|---|---|
| Detection Engineering Lab | Fixture-validated Sigma detections, target compilation, evidence manifests, and explicit telemetry-validation limits | Python, Sigma, Elasticsearch, CrowdStrike LogScale, Docker, pytest, GitHub Actions |
| ES Futures RL Trading Bot | Archived learning project retained for its Python/C# integration boundary and failure analysis—not as a trading-performance claim | Python, PyTorch, Stable-Baselines3, Gymnasium, Pandas, NumPy, NinjaScript/C# |
| NitroGen Gaming Agents | NVIDIA NitroGen fork with separately identified custom data, configuration, DAgger, and multi-frame tooling | Python, PyTorch, Hugging Face tooling, OpenCV, Windows game integration |
| Penetration Test Reports | Five authorized TryHackMe lab reports in TCM/PTES structure with per-finding MITRE ATT&CK and NIST 800-53 references; room-answer flags redacted at the text layer; a CI audit pins every PDF's hash and indicator counts | Python, pypdf, PyMuPDF redaction, GitHub Actions |
| Area | Evidence |
|---|---|
| Application engineering | Go, chi, pgx, TypeScript, Next.js, Python, FastAPI, REST, OpenAPI, OIDC, RBAC |
| Data engineering | Python, SQL, PostgreSQL, Prefect, Pandera, Splink, dimensional modeling, SCD2, ETL, identity resolution |
| Streaming and operations | Kafka, Redis, WebSockets, Prometheus-compatible metrics, OpenTelemetry traces, structured logging |
| ML and experimentation | PyTorch, two-tower retrieval, matrix factorisation, DuckDB, PyArrow, preregistered hypotheses, paired bootstrap, Stable-Baselines3, Gymnasium, spaCy, sentence transformers, BM25, reproducible evaluation |
| Security engineering | Sigma, Elasticsearch, CrowdStrike LogScale, threat modeling, secret scanning, dependency auditing |
| Cloud and delivery | AWS-oriented Terraform and Redshift designs, Docker, Kubernetes manifests, GitHub Actions, GitHub Pages |
| Quality | Unit, integration, contract, end-to-end, accessibility, performance, data-quality, and evidence-drift tests |
AWS and Kubernetes work is presented at the boundary actually demonstrated: local or CI validation and deployment design, not a paid production deployment. Portfolio soccer/member records are synthetic, and none of these projects is affiliated with or endorsed by U.S. Soccer or any member organization.
- MS Software Engineering, AI specialization
- BS Software Engineering
- AS Web Development
- MS Sport Management
- BS Parks, Recreation & Tourism Management
- AWS Certified Machine Learning - Specialty and AWS Certified Cloud Practitioner
- CompTIA Data+, Security+, PenTest+, A+, and Project+
- Microsoft Certified: Azure Fundamentals
- Practical Junior Penetration Tester (TCM)