Data systems for people who have to act on them.
Independent developer building investigative research tooling and public-interest technology, where public-records data, social justice for marginalized communities, data infrastructure and edge computing overlap.
- INVESTIGATIONS
- 11
- SOURCED FINDINGS
- 3688
- RESEARCH PAPERS
- 5
- CLIENT WORK
- 7
- PROJECTS
- 11
- TOOLS
- 28
Background
Most of my projects involve processing large volumes of public records like IRS filings, corporate registrations, and campaign finance data, then making them queryable and useful for researchers and journalists who need to understand complex networks.
I work across a wide range of stacks. I've shipped projects on Cloudflare, AWS, Google Cloud, Netlify, Vercel, and self-hosted open-source infrastructure, with knowledge graphs in Neo4J, columnar analytics in ClickHouse, and content managed in Payload, WordPress, Wix, or Shopify depending on what the client already runs. The choice of platform follows the problem, not a default.
Before software, I spent four years in financial crimes investigation at Wells Fargo, working fraud and claims cases and then supervising a team of investigators. After that I spent roughly a decade running my own e-commerce and production brands, where I was the developer, the marketer, and the person packing the boxes. That combination is why I think data systems should make complex information accessible to the people who need to act on it, not just store it.
Based in New York. Open to consulting on investigative data infrastructure, research platform builds, and public-records data tooling.
- FINANCIAL CRIMES
- Wells Fargo, 4 years
- ROLE
- Investigator, then supervisor
- OWN BRANDS
- E-commerce and production, ~10 years
- BASE
- New York City
Three areas of work
Investigative data platforms
Public-records processing at volume: IRS filings, campaign finance and corporate registrations made queryable for researchers and journalists.
Entity resolution and graphs
Resolving who is who across filings, then drawing the network. Knowledge graphs for the connections, columnar analytics for the corpus.
Edge and research tooling
Research platforms and small tools built so someone else can run them: the platform follows the problem, not a default.
Working principles
Sourced or unpublished
A claim goes public with the document it rests on attached, or it stays a draft. Corrections are published as corrections.
Raw data stays backstage
The full corpus is worked in a backend datastore; only refined conclusions reach the public layer, and nothing flows back upstream.
Fail loudly
A broken record throws at build time instead of rendering a placeholder. Every figure on this site is counted from a record, not styled in.