I build the systems that move a business's data around and act on it — pipelines, dashboards, integrations, AI tooling, and the automation that removes manual work. Twenty years across enterprise and small business: a wildfire map serving 1M+ people a month, Fortune 500 marketing intelligence, and storefronts that take payments. Available for project work.
Lakehouse and warehouse pipelines, schema design, and migrations run so they are reversible and tracked — not applied by hand and hoped over.
Scheduled pipelines, ingestion from messy third-party sources, and the monitoring that tells you when a feed goes quiet before a stakeholder does.
Dashboards and geospatial interfaces people actually use — including a live wildfire map serving over a million users a month.
Connecting systems that were never designed to talk: payments, CRM, ERP, analytics, and government or vendor data feeds.
Agent infrastructure, retrieval and embedding systems, and LLM workflows built to be evaluated rather than demoed.
Storefronts and content platforms on Shopify, WordPress, and custom stacks — catalog, payments, plugin and theme development, and the analytics behind the funnel.
Installable progressive web apps with offline support, mapping and geospatial interfaces, and the performance work that keeps them fast on a phone in the field.
Finding the expensive thing and removing it — a $100K/yr content system replaced with RSS pipelines, an $80K/yr integration platform built in-house, 20-minute deploys cut to 30 seconds.
Backtesting, walk-forward validation, and multiple-testing correction for trading and forecasting work — the same discipline applied to any question where being confidently wrong is expensive.
Engagements have ranged from Fortune 500 marketing intelligence to single-owner storefronts. Small, well-scoped projects are welcome.
Sole engineer responsible for WFCA's entire digital ecosystem.
Research bench for systematic strategies across equities, commodity ETFs, and crypto. ~138K lines of Python, 1,732 tests, CI-enforced validation rules. Deflated Sharpe Ratio for multiple-testing bias; 3-fold walk-forward as the promotion gate; alpha over buy-and-hold, not raw profit. Research and paper trading only — no live capital.
Automated search over a 72-architecture parameter space, then a forensic audit that invalidated its own headline result: the top strategy's +505%/yr came from indexing up to 6 candles into the future — with the peek removed it returned −77%. Rebuilt honestly to +55.7% over 4 years, then reported out-of-sample underperformance against buy-and-hold and declined to promote it.
Persistent cross-session memory for AI coding agents. Semantic + full-text hybrid search via PostgreSQL, pgvector, and local embeddings. REST API + MCP server.
Local-first AI coding agent with 10-phase autonomous delivery pipeline. Multi-backend LLM support (local GGUF, OpenAI, Anthropic, vLLM, MLX). 1024+ tests.
Built and run a live Shopify storefront selling custom apparel and home decor — theme development, product and inventory operations, payments, and the analytics behind it. A shipped commerce build rather than a demo: artofmetazen.com.
Cross-platform quality hooks for AI agents and git. Security scanning (OWASP), simplicity checks, naming conventions. Supports Claude Code, Codex, Gemini.
Open to conversations about data, automation, integrations, AI tooling, and platform work — at any scale. The form reaches my inbox directly.