I've spent the last couple of years figuring things out by actually building them — 19+ projects across machine learning, NLP, and computer vision, with security now creeping into the mix too. Currently doing a BS in Cybersecurity, and still learning mostly by shipping stuff and seeing what breaks.
I'm Muhammad Mohsin — a 19-year-old data scientist who learned almost everything the hard way: by building things, breaking them, and fixing them again. No fancy bootcamp shortcuts — just Python, TensorFlow, SQL, Power BI, and Tableau, turned into a habit of pulling real insight out of messy, real-world data.
Somewhere along the way I got hooked on predictive modeling, NLP, and recommendation systems — the kind of problems where the data doesn't just sit there, it actually tells you something useful if you clean it right and ask it the right questions. I'd rather ship a working prototype than perfect a theory on paper, and that mindset has shaped almost every project I've built.
Lately I've been pulling cybersecurity into the mix too — specifically how AI systems get attacked, not just how they get built. It's given me a bit of a defender's instinct that I'm now bringing back into my data and ML work. Still early in the journey, still building — but the projects speak for themselves.
Beyond the technical side, I'm most interested in turning ideas into products that people can genuinely benefit from. Every project begins with uncertainty, goes through countless revisions, and teaches something new along the way. That process of experimenting, improving, and solving problems is far more rewarding to me than simply reaching the final version.
What matters most to me is building software with a clear purpose. Eye-catching portfolio pieces are nice, but meaningful solutions create lasting value. If a project isn't addressing a real need or making someone's work easier, I'd rather rethink the approach than release something that doesn't have genuine impact.
Not a claim — a data point. This graph is built from the same 19 projects listed below, live in your browser — so you're already looking at a small proof of the data work described above.
A lightweight, client-side taste of my AI Agent Red Teamer project. Paste any AI prompt below — it's scanned in your browser, in real time, against six common attack-pattern categories. No data leaves this page.
The unglamorous 80% of every data project — paste any raw CSV and this profiles it instantly in your browser: column types, missing values, and a cleaned preview. The same first step behind every analysis in the case files above.
Open to data science, ML, and analytics roles — and increasingly drawn to the security side of AI systems. Available to work with teams anywhere.