Hey, I’m Tim, a full-stack software developer with 13 years of experience.
I have built all kinds of web and mobile applications, servers and APIs.
Interested in building something together? Let’s chat
Stealth AI Data Startup
This is a data platform run by AI agents that find public datasets on the web, clean them, write metadata, document query patterns, and keep them up to date indefinitely. You can ask questions in a chat, and agents answer by running SQL queries and creating charts, maps and dashboards on the fly. I’m the only software developer on staff, and I designed and built the whole thing, front to back.
- Designed an agent framework on a single unification: agents and tools share one typed signature and every invocation is one row in one table. Similarly, LLM chats and procedural workflows share the same data model and operate interchangeably. Streaming, persistence, resumability and provenance apply identically to a tool call, a nested subagent, a headless scheduled run or a workflow written in Python code.
- The agents get a real shell where they can write code, install packages and query data. Doing this securely meant adding hard isolation between chats, resource caps, and a firewall preventing access to the cloud metadata service. Workers stay fully stateless and scale horizontally while their long-running tasks survive deploys. I run the infrastructure too: OpenTofu on Google Cloud, with stateless app servers behind a load balancer and deploys automated through GitHub Actions.
- I wrote the autonomous curation agents that discover public data sources, write and register ingestion scripts, and schedule them as crons or multi-day backfills on a leased, resumable job queue.
- The system includes a data lake: Parquet on object storage queried through DuckDB over a restricted SQL surface, schema versions with a draft/apply lifecycle, semantic catalog search in Qdrant, and per-user access control and query scoping.
- Agents create generative-UI artifacts via a build pipeline that separates the network-enabled install step from the untrusted build step. The platform securely compiles agent-written frontend code into charts, dashboards or maps with live data access. Authentication passes to a cross-origin iframe for separation of client data.
Hot Page
Hot Page is my own business, a web site builder where you can write real code in a visual editor with guardrails to make it easier for beginner coders, designers and overworked freelancers looking for a fast solution.
- Designed and built a rich text editor for HTML code with drag and drop, instant saves and a live page preview— and the same thing for CSS code as well.
- Wrote full-stack TypeScript for a NodeJS backend and a React frontend. Integrations with third party services, including payment processing through Stripe, OAuth logins with Google and GitHub as well as AI providers.
- I maintain the server cluster in the Amazon cloud with Terraform and continuous integration on GitHub actions. Document updates are synced over web sockets using Redis.
- Wrote all the marketing material and the documentation for the platform, which has grown to more than 1,500 users. I also created a library of open-source web components called HotFX and their accompanying documentation.
Typetrigger
Typetrigger is a boutique social network designed to help writers practice their craft. Every six hours we send a new prompt to the community and writers respond with a short post. From there it works like any other social network: you can follow your favorite creators, like posts, comment and send direct messages.
- Built the website in React and two mobile apps in Flutter (iOS and Android), along with a Ruby on Rails API for the server. Our stack has a lot of pieces, including object storage, Elasticsearch, mobile notifications through Firebase, and subscription payments through Stripe as well as the Apple App Store and Google Play Store.
- I run the infrastructure that has kept the site running uninterrupted for the past 7 years using Kubernetes in the Google cloud. The server’s deployment is completely automated using GitHub actions and updates to the apps are pushed to app stores directly from the command line.
- Wrote complete unit tests for all three parts of the project: the web frontend, the mobile apps and the server. This has allowed us to keep the user experience bug free while making occasional updates over the years.
- We built these apps to be completely themeable so the same code could be used to build different social networks, such as UmeUme, which is a recipe writing app.
