About
I'm a finance and operations analyst who builds software. I spent a decade reconciling data across messy enterprise systems, learned data science and full-stack development, and now build the internal tools a commercial solar contractor runs its back office on. I understand the accounting well enough to know where automation should stop and ask a person.
What I'm doing now
- Building The internal operations platform at Day & Night Solar: data pipelines, QuickBooks automation, reconciliation, document extraction, and a governed AI agent workflow.
- On my own time Atlas, my local-first personal data lakehouse · a solar lab in the garage · still shipping updates to Cinemetrics.
- Writing Short lessons from what breaks: source of truth, failing closed, stale data, and keeping AI agents honest.
- Open to Full-time roles where accounting, data, and code meet: finance systems, analytics engineering, data and automation engineering, and internal tools.
The path here
I spent a decade in accounting — general ledger, reconciliations, fixed assets, AP/AR across enterprise ERPs at NTT, Curtiss-Wright, Robert Half, and Walmart eCommerce. Before that, I ran my own service business for 13 years: full P&L, payroll, state compliance, a team of six, and 10–12% YoY revenue growth through digital marketing.
The through-line has always been reconciling high-volume financial data across messy source systems. What changed is the tools — first Excel, then Python, then ML models from the TripleTen data science program (16 end-to-end projects: churn, pricing, forecasting, NLP, a CNN regression), and now full-stack from SavvyCoders. Today I design AI-assisted workflows and agent-based systems on Claude Code, build domain-specific databases and tooling, and work at the intersection of finance, operations, and automation.
Right now that means the internal operations platform at Day & Night Solar, a commercial solar and energy-storage contractor, where I'm the sole developer and still do the finance work it supports — and writing down what breaks while I build it.
How I learn
I learn best by building something real enough to break. Usually a real problem comes first. I research enough to build, use AI to shorten the feedback loop, and when the AI's explanation doesn't match what the system actually does, I go inspect the system. I make a prediction, measure what happened, and turn the difference into a rule I can reuse. I keep the original data so I can re-check an answer instead of trusting it because it sounds right.
How I work
- One fact, one owner. Reports explain the truth; they don't get to own it.
- Fail closed near money. Automation prepares the work; a person approves it.
- Unknown is a real state. Missing is not zero, and stale is not current.
- Check the grant, not the description. A claim needs evidence before I build on it.
- Controls, audit trails, and reversible operations by default.
Skills
Core — used regularly, confident Working — shipped with it, still growing Exploring — actively learning
Programming & AI
Where I spend most of my build time right now.
Data & Engineering
Turning messy source systems into something you can actually query.
Machine Learning
Classical ML where I've shipped measurable results, plus what I'm building toward.
Full-Stack Web
The SavvyCoders stack — built a real capstone in it.
Finance & Enterprise Systems
The foundation — still shapes how I think about controls, audit trails, and data integrity.
Business Domains
Where my work has made real decisions.