Work · Product engineering

Systems built for the difficult part.

The interface matters. These case studies go one layer deeper into the state, policy, platform, and trust problems that made each product difficult to build well.

Experience and skills Combined GitHub activity Contact Steven

01 · Selected Work

Built where trust is part of the architecture.

Three systems where the hard part was not the screen. It was keeping money, policy, platforms, and sensitive information coherent.

01A

Property-management financial SaaS

TEO

Lead Full-Stack Engineer

Problem: A financial product has to preserve one trustworthy account of an invoice while QuickBooks, hosted checkout, webhooks, and background jobs all change it.
What I built: I own payments and accounts receivable end to end: QuickBooks invoicing through Payabli hosted checkout, webhook processing, background workers, and invoice status derived from the ledger. I also founded the design system from zero with a three-layer token architecture and ESLint enforcement.
What made it hard: The money path must fail closed. I hardened it with default-deny safeguards, established an adversarial multi-model AI review gate, and led WCAG remediation across the product.

TypeScript
QuickBooks
Payabli
Webhooks
Background workers
Design tokens
WCAG

Read full case study

01B

Creator learning platform

Musterhall

Sole Engineer

Problem: One learning product needed to behave coherently across web, mobile, real-time rooms, and a database with product rules that could not drift by client.
What I built: I built the full platform: a Next.js and TypeScript web app, Flutter mobile app, and Supabase/Postgres backend with schema-enforced invariants. A single token source compiles to web CSS and a generated Dart package, with a public Storybook for the system.
What made it hard: LiveKit rooms required a real-time engine with host moderation. A feature-flag rollout ladder let the platform introduce that capability deliberately across surfaces.

Next.js
TypeScript
Flutter
Supabase
Postgres
LiveKit
Storybook

Read full case study

01C

Legal intelligence portfolio

CrimCaseAI

Principal Engineer

Problem: Legal teams need useful intelligence from sensitive case material without weakening privilege, provenance, or the operating boundaries of restricted environments.
What I built: The portfolio includes a live case-management SaaS, a generation-two rebuild with schema-generated forms and in-browser ML, and an evidence-intelligence platform serving paying law firms.
What made it hard: Aegis extends the same work into an on-prem Docker appliance: multimodal ingestion, local OCR and transcription, vector search, entity graphs, and citation-gated AI chat, with privileged data kept on hardware the firm controls.

In-browser ML
Docker
Local OCR
Transcription
Vector search
Entity graphs
Grounded AI

Read full case study