I build Python tools that make data easier to trust and software easier to understand. My focus: machine learning, data quality, and interfaces that let you explore what the code is doing.
Based in India · Open to data science / ML internships
What happens when the data changes? A reproducible ML experiment and drift-monitoring dashboard. Compare models, explore synthetic shifts, or bring your own numerical CSVs.
|
Catch bad data before the pipeline does. Catch broken CSV data in GitHub Actions. Validate contracts, keep downloadable reports when a check fails, and explore results offline.
|
Break the code. See why. Test the fix. Step through concurrency bugs one operation at a time. Rewind, try a different schedule, and share the exact execution that broke.
|
From a reported issue to a recorded fix. A maintenance-workflow demo with private workspaces, asset maps, photo reports, public tracking, and an audit trail.
|
Also on my workbench: Python Algorithm Lab—readable algorithms with runnable examples and independent tests.
I investigate edge cases, reproduce failures, and submit focused fixes. These are submitted upstream pull requests; each link shows the current review and test status.
| Project | Proposed fix | Evidence |
|---|---|---|
| The Algorithms · Python | Keep power iteration from stopping early on negative dominant eigenvalues. | Reproduction, numerical checks & review |
| The Algorithms · JavaScript | Find existing keys in singleton and constant arrays with interpolation search. | Fix, regression tests & CI |
| keon/algorithms | Avoid overflow and underflow when computing cosine similarity for finite vectors. | Scaling fix & regression tests |
| Data & ML | APIs & interfaces | Testing & delivery |
|---|---|---|
| Python · NumPy · pandas | FastAPI · SQLAlchemy | pytest · Ruff |
| scikit-learn · SciPy | PostgreSQL · SQLite | Git · GitHub Actions |
| Jupyter · Matplotlib | React · TypeScript · Tailwind | Reproducible experiments |
I also edit short-form and sports videos in DaVinci Resolve. Code and a timeline are two different ways to turn a rough idea into something people can use—or watch.
See my video work ↗ · Visit my portfolio ↗
A little more: engineering details & GitHub activity
Some smaller changes show how I approach the edges of a project:
- Protecting input files and making report writes atomic, with failure and file-alias regression tests.
- Keeping missingness alerts consistent through export and reimport.
- Rejecting malformed inventory updates without changing saved state.
Have an internship, an interesting dataset, or a project to talk about? Let's connect.