I'm a fourth-year Computer Engineering student at Hormozgan University. Most of what I do these days circles back to AI in healthcare — I like problems where the stakes are real and the data is messy. Alongside that, I work as a teaching assistant, which has probably taught me as much as it's taught the students.
Working with my advisor on detecting and segmenting brain tumors from MRI scans. The interesting part isn't getting a model to work in general — it's getting it to work on the small, oddly shaped tumor regions that usually get lost in the averages. I'm comparing a few segmentation approaches, from U-Net variants to newer transformer-based ones, to see what actually holds up there.
I put together and grade the algorithm exercises for the course, and spend a fair bit of time in office hours just talking students through where their logic is going sideways — usually more useful than any amount of grading.
Similar role here — assignments around logic and set theory, plus a weekly session where people can bring whatever's not clicking. It's a good way to find out what's actually confusing versus what just sounds confusing in lecture.
Built a CNN from scratch to classify brain MRIs into the four stages of Alzheimer's. The dataset was small, so most of the actual effort went into keeping the model from just memorizing it.
Scraped and cleaned listings off Divar, then tried a handful of models — mainly XGBoost — to see how well location and property features could actually predict price in a market that doesn't always behave rationally.
A sentiment classifier for Amazon reviews, partly as an excuse to get more comfortable with NLP preprocessing and a couple of transformer-based approaches I hadn't used much before.
Trained a model to recognize Persian traffic signs, with extra preprocessing to handle the kind of low-light, low-contrast images you actually run into on real roads rather than in a clean dataset.
English — Fluent · Persian — Native
AI in healthcare, machine learning, and research more broadly.