Work examples

Work and demos, with honest availability labels.

This page reframes the demos as evidence for employers and technical decision-makers. Each example shows what it proves, where the safety boundary sits, and whether the work is public, protected, prototype, or internal-only.

Production / public

Public surfaces that can be reviewed directly, with claims limited to what is actually available.

Protected demo

Interactive surfaces behind access control, email allow-listing, and manual approval where risk requires it.

Prototype / internal-only

Architecture, workflow, and outcomes are public; private systems, credentials, customer data, and internal infrastructure stay private.

Proof hierarchy

The point is credibility, not demo volume.

Clinical AI is the strongest employer signal. Operations and MRO show workflow translation. Trading, security, and music add range, data scale, boundary-setting, and product proof.

Every card states what the work proves, not just what it is.

Status labels distinguish production, protected demo, prototype, and internal-only work.

Sensitive domains use synthetic data, access controls, and conservative public boundaries.

Work examples

What each system proves.

The cards are written for evaluation: value proposition, employer proof, and safety or access boundary.

Protected demo

DRG Clinical Validation

Coder-grade DRG validation built around an extraction -> normalization -> validation -> human-review loop with explicit guardrails.

What it proves

Unstructured clinical inputs become structured, auditable outputs with a human-in-the-loop gate before anything moves downstream.

Safety / Access boundary

Live demo behind Cloudflare Access. Email allow-list, no patient data. Manually reviewed within 24 hours.

Production

TwoSevens Music

AI-assisted artist catalog and karaoke-style listening demo, with stem-extraction studio tooling kept internal until it is ready to be public.

What it proves

End-to-end music surface: hosted catalog, public listen demo with karaoke-style lyrics, and a credible bridge from catalog into SongCraft commerce.

Safety / Access boundary

Public catalog and listen demo are live. Stem extraction is internal studio tooling. No public commerce, payouts, or admin panels are exposed.

Prototype

SongCraft

Personalized-song commerce MVP - order intake, Stripe checkout, AI-assisted generation, human review, and MP3/lyrics delivery for the occasions that deserve a track.

What it proves

Generative-AI commerce loop built end to end with explicit review gates between request, payment, fulfillment, and delivery - not a one-click black box.

Safety / Access boundary

MVP is code-complete and merged, running manual fulfillment in Stripe test-mode staging only. No public checkout and no live payments - production launch stays gated until the launch-readiness criteria are signed off.

Prototype

RegenITAD - MRO Asset Listing

AI-assisted listing pipeline for ITAD and surplus IT equipment: extraction, vendor enrichment, normalization, and review before any draft is exported.

What it proves

Workflow-gated state with evidence capture and compliance-aware document generation. AI accelerates, humans approve.

Safety / Access boundary

Planned MVP. Sample / synthetic assets only - no real customer asset tags, no direct marketplace publishing in MVP.

Prototype

Opportunity Evaluation

AI-assisted opportunity evaluation and validation: extract requirements, validate fit, identify gaps and risks, score evidence, and prepare reviewer-ready next steps.

What it proves

The same criteria-driven workflow can evaluate jobs, RFPs, vendors, partnerships, sales leads, and other opportunities without becoming an auto-apply bot.

Safety / Access boundary

Public proof uses synthetic/sample opportunity data only. No real resumes, recruiter emails, job applications, private documents, or personal data are shown.

Protected demo

Security Platform

Controlled white-hat research sandbox with real-time monitoring, anomaly detection, and compliance reporting.

What it proves

Defensive posture and reporting hygiene - alerting paths, evidence trails, and reviewer workflows wired up before any tooling.

Safety / Access boundary

Authorized use only. Lab/synthetic targets only. No live offensive tooling, internal targets, credentials/secrets, or public scanner access.

Internal-only

Trading / ML Platform

Simulation-only ML research surface for historical market data, feature engineering, model workflow experimentation, backtesting, and risk-control review.

What it proves

Production ML lifecycle discipline: historical/synthetic data ingest, feature store, experiment tracking, replayable backtests, and explicit risk controls.

Safety / Access boundary

Historical/synthetic analysis only. No financial advice, live broker execution, credentials, account balances, or order routing are exposed.

Internal-only

Infrastructure Lab

Distributed operations across ~25 servers with operator-controlled orchestration, monitoring, and recovery discipline.

What it proves

Multi-node operation is treated as a real product surface - service separation, queue-backed work, and recoverability are first-class.

Safety / Access boundary

Architecture and outcomes only. No internal hostnames, IP ranges, admin URLs, or credentials are published.

Reference protected demo

DRG Clinical Validation

Coder-grade DRG validation built around extraction, normalization, validation, and human review. The live interaction is gated behind Cloudflare Access, uses synthetic / non-PHI examples, and is manually reviewed before access is granted.

Available for evaluation

Need a focused evaluation path?

Start with the work examples, then open a hiring or workflow conversation. Protected demos stay gated; public pages keep claims conservative.