I built Audio Transcriber, a lightweight browser-based app that converts audio recordings into downloadable text files using a local Whisper server. This project was created to solve my own need for taking accurate notes during conversations with doctors and caregivers. After exploring cloud and desktop options, I designed a local web app powered by a Python/FastAPI backend and a clean Svelte frontend, allowing fast transcription, zero cloud costs, easier handling of large files, and full privacy on my machine. Along the way I learned FastAPI fundamentals, built a modern Svelte store architecture, set up a clean dependency workflow with requirements.txt, and tied the project together with a simple start.sh script. This open-source tool has already improved my daily workflow, and the codebase is available for anyone to use or extend.

For fun and as a way to explore cutting-edge technology, I built and deployed bretteastmanstudio.com, a site hosting my music teaching materials alongside a music historian chatbot for students seeking deeper knowledge about artists and topics. I integrated Google's lightweight gemini-2.5-flash model with contextual memory of the past 5 chats to deliver functionality at minimal cost. The backend runs on a cost-effective Linode Nanode instance hosting an open-source Pocketbase database, secured behind an NGINX reverse proxy with HTTPS via Let's Encrypt and rate limiting for protection. The project leverages SvelteKit, TypeScript, Tailwind CSS, and Cloudflare, demonstrating my ability to architect full-stack applications that combine modern web technologies with AI integration and robust deployment practices.

As a volunteer Full Stack Software Engineer at SF Civic Tech, I was on a team of developers and designers working on Project Compass, a web application to streamline the process of setting and recording Individual Educational Plan (IEP) data for students with disabilities, for the San Francisco Unified School District (SFUSD). I was responsible for implementing a feature to the project, using Nodemailer, to automatically send confirmation emails to new para-professionals as they are added to the database by the teacher. I also engineered the design team's designs for the Student page and Staff page, implementing the backend CRUD procedures as well.

The Ranked Choice Voting Calculator is a single page web application designed to facilitate the tally of a ranked choice vote involving 3-5 candidates and up to 20 voters. I originally came up with this idea to solve a problem. My friends and I have a music listening group where we use ranked-choice voting to decide what album we will listen to together. I chose Svelte as the framework because of its small bundle size and fast load times. I used Vite as the bundler because of its speed and ease of use. I also used Tailwind for styling because of its utility-first approach.