Case Studies

Systems I designed — and the judgment behind them.

I design AI-augmented systems, direct the AI that builds them, and verify them end-to-end. Each case study below walks through one real system: the flow, the controls, and the calls I made — not a feature list, the reasoning underneath.

SmartQuote ProDesigned · Directed AI build · Verified end-to-end

An AI-augmented quoting & margin-review system I designed and operate for a facilities-maintenance business.

Intake → priced quote → risk-gated human review → audit trail. The margin control is built into the pricing logic, role-scoped access is enforced server-side, and an LLM drafts the strategy memo but never makes the decision.

Read the case study →

JumpOnionDesigned · Directed AI build · Verified end-to-end

A computer-vision + rule-engine system that diagnoses figure-skating jumps — measurement first, LLM as translator.

Deterministic biomechanics measure the jump; an LLM only translates the verdict into language. Rotation-defect calls that need a human are routed to expert review, never auto-displayed.

Read the case study →

More case studies in progress

IvyBloomIn progress

An adaptive-learning platform with AI cost discipline and mastery tracking I designed end-to-end.

Multi-AI lesson generation with refund-on-failure on paid routes, assessment-calibrated mastery tracking, and a portfolio builder that turns months of learning into application-ready records.

SunnyInvoicesIn progress

A three-layer automation that collapses a repetitive manual billing process into one click.

A browser-script + Node + Playwright pipeline that turns a 5x manual workflow into a single action — built to solve a real operational pain, not as a product.

These are working systems, shown through synthetic, debranded data — no real client, pricing, or private information appears on any page. The goal is to show how the systems are designed, not to expose what runs inside them.