Final-year CS undergrad. I'd rather write the tokenizer than import it — most of my work sits close to the metal, and I'm headed toward research in embodied AI.
tokenizer — a BPE tokenizer in Rust with Python bindings, built to match OpenAI's tiktoken on cl100k_base / o200k_base and run faster.
CodeNeuron — a deployed ML platform I co-built. I own the backend, the execution isolation, and the cloud infra: FastAPI, Postgres, Docker, Cloud Run, CI/CD.
Vision — object detection and segmentation, PyTorch.
Next: a RAG system built around evaluation rather than the demo — hybrid retrieval, grounded citations, and an eval harness that actually measures whether it works.
Cinema between training runs.