ML/AI engineer, Kaggle competitor, and open-source contributor.
I contribute real code fixes across the scientific-Python and AI ecosystem — the method is always: reproduce the bug locally, ship the minimal fix with regression tests, respond to review.
Merged: psycopg, OpenCV (×3), scikit-image, Hugging Face diffusers
In review: dask, pygments, apache/libcloud, google-cloud-python, PyInstaller, pandas, scikit-learn, PyTorch Lightning, SymPy
The day-by-day ledger of bugs found, fixes verified and reviews handled lives in open-source-journey.
- spanguard — span-level prompt-injection guard for AI agent tool traffic: strip the attack, keep the data. Fine-tuned open-weights span model with committed benchmarks against Llama Prompt Guard 2.
- IGNIS — NASA Space Apps Challenge 2026: a 26-year harmonized MODIS+VIIRS fire-activity calendar with early warning, in-browser and key-free.
- arc-agi-2-solver — ARC-AGI-2 solver: Qwen3-4B NVARC+ with test-time training.
- solar-filament-slm — FilaNet: a 4.5M-parameter solar filament instance segmenter (GONG H-alpha, Panoptic Quality metric).
- k12-misconception-diagnoser — LoRA-based diagnosers for K-12 math misconceptions (Kaggle).
- kaggle-2026-competition-lab — research, rank-1 strategies and starter kits for active Kaggle competitions.
ARC Prize 2026 (35.53 final private-board score, full writeup) · NFL Big Data Bowl 2027 · Solar Filament Segmentation Challenge 2026 · Misconception Detection in K-12 Student Math Workings


