Skip to content
View Hisernberg's full-sized avatar
🏠
Working from home
🏠
Working from home
2 followers · 7 following

Achievements

Achievement: Pull Sharkx3Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: Pull Sharkx3Achievement: YOLOAchievement: Quickdraw

Block or report Hisernberg

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Hisernberg/README.md

Hi, I'm Nabid Nur (Hisernberg)

ML/AI engineer, Kaggle competitor, and open-source contributor.

Open source

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 (17 upstream PRs): psycopg, OpenCV (×3), scikit-image, Hugging Face diffusers, dask, apache/libcloud, nats.py, coqui-ai-TTS, pypdf, nibabel

In review (~115 open): pygments, google-cloud-python, PyInstaller, srsly, cheroot, PaddleOCR, falcon, twisted, behave, lxc — plus pandas, scikit-learn, PyTorch Lightning, SymPy. Dead-link fixes are verified 404 with live replacements checked before opening; deprecation fixes preserve naive-datetime behavior exactly.

The day-by-day ledger of bugs found, fixes verified and reviews handled lives in open-source-journey.

Featured projects

  • 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.

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

Popular repositories Loading

  1. spanguard spanguard Public

    Span-level prompt-injection guard for AI agent tool traffic: strip the attack, keep the data. Built for Meta Global AI Hackathon 2026 — fine-tuned open-weights span model, committed benchmarks vs L…

    Python 3

  2. ai-agent-projects ai-agent-projects Public

    ai agent models with groq

    Jupyter Notebook 1

  3. capstone-project-java-notebookllm capstone-project-java-notebookllm Public

    1

  4. Decode_labs-Artificial-Intelligence_internship--task--3 Decode_labs-Artificial-Intelligence_internship--task--3 Public

    Jupyter Notebook 1

  5. arc-agi-3 arc-agi-3 Public

    Python 1

  6. tracea tracea Public

    Trace anything running - dynamic reverse engineering for AI agents. MCP server + CLI with evidence records: hooks, syscalls, memory, network. The dynamic half of REA.

    1