CS + AI/ML student
AI/ML and systems builder interested in reliable software, thoughtful evaluation, and practical research tools.
I work across LLM agents, reinforcement learning environments, evaluation systems, AI infrastructure, and applied engineering. This site collects my projects, writing, publications, and research notes as they develop.
Profile
I like problems where models, tools, data, and systems meet, especially when behavior needs to be measured carefully rather than only demoed.
I use this space as a working record of what I am building and learning: research prototypes, software projects, implementation notes, and longer-form writeups. The common thread is a preference for systems that can be inspected, tested, and improved over time.
Interests
A few areas I keep returning to while building projects and reading research.
Current research thread
A mechanically verified infrastructure-diagnosis RL environment for training agents to repair faults through bounded JSON actions, executable sandbox rewards, and state-based checks instead of LLM judges.
Writing
Research diary entries, project notes, and implementation writeups.
A deeper look at CrashDiag, the first Indium AI Labs environment for training agents to diagnose and repair infrastructure faults with executable rewards instead of LLM judges.
Why I am building CrashDiag and starting Indium AI Labs to create mechanically verified reinforcement-learning environments for infrastructure and security agents.
Why TokenScope matters for CrashDiag, what shipped in v0.1.0 and v0.2.0, and how tokenizer inspection becomes part of reliable agent evaluation.