Most finance analysts can't build software, and most people who build software don't know finance. I do both. On one side: budgets, forecasts, variance analysis, month-end close, three-statement models. On the other: code, automation and AI agents. I use the second to make the first faster, clearer and harder to get wrong. Here's my work.
For the past four years I've been a financial analyst at a Bay Area housing services organization, with full responsibility for a portfolio of 25 programs: the annual budget, the year-end forecast, the monthly variance analysis, the month-end close, and billing our funders every month. I also build the tools we do this work with.
When my monthly variance analysis flagged a projected overspend early, I recommended moving money across programs; leadership adopted it and avoided a mid-year gap. My year-end forecasts showed funding shortfalls before they hit, so leadership could adjust hiring and funding in time.
Two of the systems I built are below, rebuilt here on made-up data; the real numbers are confidential.
The live dashboard our monthly budget review runs on, rebuilt here on made-up data. It shows how I turned an Excel workbook pasted in by hand into a live system that leadership and program managers use every day.
Open the dashboard
The Power Query and Python pipeline behind our month-end invoicing, rebuilt here on made-up data. It shows how I redesign a process instead of automating the old one: one ledger export, one staffing file, and a check on every invoice before it goes out.
Open the workbookOutside my job: models built on public filings, public data and one made-up company, and two products of my own.
A driver-based annual hiring and spending plan for a made-up software company of about 300 people, with a one-page recommendation memo. It shows headcount and opex planning, scenario and sensitivity work, and writing a decision for executives.
Open the plan
A month-end variance pipeline for a made-up software company: it computes every variance against plan, a model drafts the commentary, and checks tie each number back to the books. It shows automating the close with controls, and knowing when the machine is wrong.
Open the pipeline
A quarterly three-statement model of Toast, the restaurant software company, built from its SEC filings, with scenarios and checks. It shows how I break a business into the drivers that decide its profit and forecast them.
Open the model
A two-page decision memo on whether Meta should keep funding Reality Labs, with the Excel model under it, built from Meta's SEC filings. It shows building an investment case: payback, scenarios, and the conditions that would make me stop.
Open the memo
A pricing case on HubSpot's 2024 move to per-seat pricing: a memo and an Excel model on how to migrate its existing customers, built from its filings and archived pricing pages. It shows pricing work: sizing revenue at risk, testing price sensitivity and setting rollout and rollback rules.
Open the case
A unit-economics case on Duolingo's move of its AI conversation feature into the Super tier: a memo and an Excel model that price each call on archived API price lists. It shows compute-cost modeling, usage-based packaging and reconciling a model to a company's own disclosures.
Open the model
A benchmark of what 124 US-listed software companies spend on sales and marketing, R&D and G&A, built in SQL from SEC filings. It shows working with large, messy public data and turning it into planning targets.
Open the benchmark
A subscription revenue ledger rebuilt in SQL from a public SaaS dataset, with the MRR bridge and deferred revenue tied to the cent. It shows how I clean messy revenue data and make it reconcile.
Open the ledger
Real-time risk-management software for futures traders that I founded in April 2026 and run myself, listed in the NinjaTrader Ecosystem. It shows owning a whole product: engineering, pricing, affiliates, marketing and support.
Read about Meridian
A Windows app and Excel add-in that turns Excel files into print-ready PDFs in one batch, live and free to download. It shows building and shipping a small product end to end.
Visit crazysmartpdf.comFinancial analyst at a Bay Area housing services organization since 2022. B.A. in Economics, UC San Diego, 2022. Founder of Meridian since April 2026.
I run an AI agent system that writes code and workbooks to my spec. Every project on this site was built that way: I set what it had to do, reviewed every result, and a separate check recomputes the figures from the source data. The two from work are rebuilt on made-up data; the real numbers are confidential.