About
I work in the open, and the receipts are in the repo.
Final-year Computer Science student at NED University in Karachi, specialising in AI.
The short version
Eight weeks, raw rows to a deployed paper.
I spent eight weeks taking a real production search dataset from raw rows to a deployed research paper: data contract, frozen baseline, model, validation, write-up.
The hardest part wasn’t any single technical wall. It was finishing, alongside a full course load, without quietly dropping the tests I’d promised myself I’d run.
The paper reports five archetypes instead of six because one of them didn’t survive those tests.
I work in the open. Every number in the paper is regenerated by a committed notebook, and the receipts are JSON files in the repo, so nothing rests on my word.
Tools
What I reach for, and what I logged.
- Python
- pandas
- NumPy
- scikit-learn
- DuckDB over Hugging Face
- matplotlib
- Git
- GitHub Pages
- Jupyter
- Colab
I used Claude throughout, and logged how: the prompt-iteration record is in the repo alongside the notebooks.
Looking for
A first full-time role, remote.
A first full-time data or ML role, remote.
If you’d rather read someone’s notebooks than their CV, the repo is open and the paper is live.