Senior Full-Stack Engineer · AI Automation
I build systems that keep working after everyone logs off.
Senior full-stack engineer and AI automation specialist with 15+ years of continuous production experience across US, UK, and Philippine markets — resilient data pipelines, AI-assisted workflows, and the rare skill of making 20-year-old enterprise software talk to modern stacks.
76 automated tests introduced into an untested codebase. 12+ concurrency and security defects caught before merge. A Chrome extension live on the Web Store with public users. Remote · GMT+8.
Engineering, translated into business outcomes.
Every tool below has shipped to production. Each one is listed with what it actually buys a business — not just what it's called.
Backend Architecture
- Ruby on Rails & Django (15+ years, production) — Battle-tested platforms that scale for a decade, not a demo: one migration I led kept a single production system evolving for 10 straight years.
- Service Object Pattern + atomic transactions — Zero-drift financial operations: every multi-step money movement completes fully or not at all, so ledgers never silently disagree.
- Idempotency-key design — Duplicate-proof workflows: retries, double-clicks, and network hiccups can't double-post a transaction or double-charge a customer.
- PostgreSQL & MySQL schema design — Databases modeled for the queries the business will run in year three, not just the demo in week one.
- Multi-tier fallback pipelines (live scrape → cached snapshot) — Near-100% uptime against unreliable upstream sources: when a third-party site breaks, the dashboard stays populated instead of going dark.
AI & Automation
- AI API integration (Claude, OpenAI, Gemini, OpenRouter, Ollama) — Multi-provider AI behind one interface, so a product is never hostage to a single vendor's pricing or outage.
- Automated content generation pipelines (Puppeteer + FFmpeg) — 2,016 rendered videos from one pipeline at zero marginal cost per unit — content output decoupled from human hours.
- Prompt engineering + defensive output handling — AI features that degrade gracefully: malformed model output gets caught and corrected instead of crashing the workflow.
- Production web scraping (Playwright, Cheerio) — Structured data extracted from sources with no public API — market intelligence competitors have to collect by hand.
- Scheduled background jobs with stale-only refresh — Compute costs bounded to exactly the data that changed, not the whole catalog on every run.
Full-Stack Delivery
- Next.js, React, TypeScript — Modern, typed frontends shipped end-to-end — from data model to deployed UI — without a handoff between "backend person" and "frontend person."
- Chrome Extensions (Manifest V3, published to the Web Store) — Confirmed shipping capability from build to store approval: ReplayHub is live with public users, a second extension built on the modern declarativeNetRequest standard.
- Cross-platform desktop apps (Python/Qt) — One codebase serving Windows, macOS, and Linux — a third of the maintenance surface of three platform-specific builds.
- Zero-dependency static delivery — A production site rebuilt to under 20 KB total page weight with no build tooling — instant loads, nothing to patch, nothing to break.
- Automated testing discipline (pytest, Vitest) — 76 tests built from zero coverage on a previously untested GUI app, run headlessly in CI — regressions caught by machines, not customers.
Legacy & Enterprise Integration
- QuickBooks Desktop integration (QODBC + pyodbc, 32-bit Python) — A genuinely rare skill: bridging 20-year-old desktop accounting software to modern web platforms without ripping out the system the business runs on.
- Double-entry accounting systems — Financial modules where debits always equal credits by construction — validation the auditors can verify, not take on faith.
- Stripe production billing (subscriptions + webhooks) — Recurring revenue infrastructure that has processed real customer payments, including launching an entirely new revenue stream for a healthcare SaaS.
- Monolith → JSON API migration — Incremental modernization of a live platform over a 10-year lifecycle — no big-bang rewrite, no downtime, no lost customers.
- Elasticsearch & Kibana — Enterprise search and analytics run in production, turning millions of records into answers in milliseconds.