Independent Researcher | 18 y/o | RC hobbyist
I build across developer tooling, machine learning research, and music.
Right now, most of my work sits in three buckets: systems tools for real developer workflows, research on parameter-efficient deep learning in small-data settings, and public notes or papers that document what I am learning as I go.
- Building practical CLI tooling and systems software with a focus on ergonomics and real-world use.
- Researching small-data, parameter-efficient neural architectures for materials-property prediction.
- Publishing papers, preprints, and open research notes around AI, compression, and adjacent topics.
| Project | What it is |
|---|---|
| Munchkin | Cross-platform multi-account manager for Codex, Gemini, and Qwen CLIs with encrypted storage, fast switching, quota tracking, and Codex load balancing. |
| TRIADS | Recursive attention architecture for materials-property prediction in small-data regimes, with code, pretrained weights, benchmark results, and a linked preprint. |
| YouTube TLDR | Chrome extension that redirects YouTube viewing into an AI-generated summary first to reduce distraction and force a deliberate choice. |
| Meow-Mocks | Web-based mock test platform driven by JSON exam definitions, with timed exams, review flow, and results analysis. |
| CuriosityCache | Public notebook for research notes on AI, compression, neuroscience, transformers, and related questions. |
- TRIADS: Tiny Recursive Information-Attention with Deep Supervision Preprint on Zenodo, March 24, 2026.
- Why "Infinite" Lossless Compression Fails Paper on Zenodo, 2026.
- Portfolio: rtx09x.github.io
- ORCID: 0009-0004-8521-7627
- LinkedIn: rudra-tiwari-b275753b5
- GitHub: @Rtx09x
- YouTube: @rtxmusicx9
- Spotify: Artist Page
- I am interested in first-principles thinking, scientific curiosity, and building things that are concrete enough to test.
- I do not treat research, engineering, and creative work as separate boxes; most of my best work comes from the overlap.