<!-- /projects.md: the projects page, as plain markdown -->

# Projects

Seven things I’ve built, newest first.

## OneStar

2026. Audience research.
Audience sentiment for apps and games, compared across the App Store, Google Play, and Steam and tracked as new reviews arrive. Jev, a small model that returns typed answers instead of text, reads every review and code does all the counting.
Stack: Python, FastAPI, Jev, Server-sent events, PostgreSQL, TypeScript, React, Next.js, Fly.io
Areas: AI routing systems, API design, Backend systems, Full stack development
- [Website](https://onestar.sh)

## Pyx

2026. Desktop app.
A local-first portfolio tracker and research terminal. Every account lands in one encrypted file on your machine, with charts, research, notes, and an MCP server for the AI of your choice.
Stack: Rust, Floem, wgpu, tokio, SQLite, SnapTrade, Tantivy, React, Tailwind CSS
Areas: AI retrieval systems, API design, Backend systems
- [Website](https://pyx.finance)

## Cure AI

2024. Research platform.
A scientific research platform with access to 26 million pieces of peer-reviewed literature. Researchers ask questions in plain language and every answer links to its sources. Founded during my junior year at Emory. The retrieval method was granted a patent, and the product was named Product of the Day on Product Hunt.
Stack: Python, FastAPI, React, Tailwind CSS, Pinecone, OpenAI, Stripe, Neon, Vercel
Areas: AI retrieval systems, API design, Backend systems, Full stack development
- [Website](https://www.askcureai.com)

## U.S. Wildfire Analysis

2024. Data science.
An analysis of 1.88 million U.S. wildfire records from 1992 to 2015, examining where large fires cluster, when they occur, and how long they take to contain. Gradient-boosted trees produced the most accurate containment-time predictions.
Stack: Python, pandas, NumPy, scikit-learn, Matplotlib
Areas: Data science and ML
- [GitHub](https://github.com/Jared-Watson1/CS470)

## Loan Default Prediction

2024. Machine learning.
A study of fairness in automated loan approval, training three models on 890,000 Lending Club applications. Removing bias-related features had minimal effect on accuracy, suggesting fairer lending decisions are achievable without a meaningful tradeoff in performance.
Stack: Python, scikit-learn, pandas
Areas: Data science and ML
- [GitHub](https://github.com/Jared-Watson1/loan-default-prediction)

## DooleyAFavor

2023. Full stack.
A task marketplace for Emory University students to post jobs and hire peers. Developed with a small team using sprint planning, with a Flask REST API, a PostgreSQL database, and deployment on Heroku.
Stack: Node.js, Python, Flask, PostgreSQL, Heroku
Areas: API design, Backend systems, Full stack development
- [GitHub](https://github.com/Jared-Watson1/DooleyAFavor)

## Dodge

2022. Game.
A 2D game in which enemies calculate intercept trajectories rather than following the player directly, so difficulty scales with skill. Built independently in Python with Pygame, including custom collision detection and a particle system.
Stack: Python, Pygame
- [GitHub](https://github.com/Jared-Watson1/Dodge)

---
More: [Home](/index.md) · [Projects](/projects.md) · [Blog](/blog.md) · [GitHub](https://github.com/Jared-Watson1) · [LinkedIn](https://www.linkedin.com/in/jared-watson-b7b5b6220/)