I'm a software engineer who builds things with AI and occasionally remembers to contribute back to the OSS projects I shamelessly depend on.
Not another ChatGPT wrapper, I promise. Well, not just wrappers:
- AI productivity tools - Because apparently we need AI to be productive now
- Developer automation - Teaching robots to do the boring stuff since 2025
- FinOps tooling - Someone has to watch those OpenAI API bills spiral
- OSS contribution frameworks - Systematically giving back to maintainers who deserve better
Frontend: TypeScript, Next.js, React (yes, I know Svelte exists) Backend: Python, Node.js, PostgreSQL (the boring reliable stuff) AI/ML: PyTorch, Transformers, OpenAI, Anthropic, Mistral (collecting API keys like Pokémon) Infrastructure: Docker, AWS, Vercel (pray the serverless cold starts are merciful)
Turns out you can actually contribute to the libraries you use instead of just starring them:
- Keras - Fixed docs because reading is important (#21827 ✓ merged)
- PyTorch - Type annotations and code cleanup (#167241, #167256)
- Working across AI SDK repos - Because one LLM provider is never enough
Following the AI hype cycle very responsibly:
- Multi-agent systems - Because one AI wasn't chaotic enough
- RAG architectures - Turns out models don't actually know everything
- Context windows - Remember when 8k tokens was impressive?
- Fine-tuning - For when prompts just aren't cutting it
- AI-native workflows - Whatever that means this week
- I have more API keys than passwords (this is fine)
- My
package.jsonis 90% AI dependencies - I've starred 500+ repos I'll "definitely read later"
- Contributing to OSS is cheaper than therapy
Where the BBQ is fire and the internet is... adequate.
"In a world full of GPT wrappers, be the person who actually reads the docs."
Last updated: Nov 2025 • Built with TypeScript • Deployed on Vercel • Runs on caffeine