ABOUT ME

I am curious

I earned my bachelor's from K. N. Toosi University of Technology, where a strong drive for practical skills led to several projects I'm proud of, and then completed a master's in Electrical Engineering at Sharif University of Technology, where I explored AI and built a solid foundation across its branches. Now, as a Ph.D. student at DTU, I focus on deep learning and TinyML, exploring ways to make AI models more efficient and practical for real products—enabling applications with stronger privacy, lower energy use, and better performance. The best is yet to come...

Welcome to my less-than-professional website. My name is Mohammad Amin Hasanpour, and since you're here, I'm guessing you're curious about who I am. So, let's dive in!

I like to think of myself as a follower of my passion—whether that means diving into dozens of seemingly unrelated skills (like web development, which led me to build this very site) or chasing curiosities all the way into a Ph.D. If that sounds self-motivated, I'll take it (but please, no “geek” labels, OK?). And of course, passion is fun, but diligence is what makes it stick. There's a reason some people are on top :)

Oh, and about my creativity ? I'll keep it humble - no need to come off as cocky, right?

Projects

< CODE /> The art of creating

Somewhere between caffeine and curiosity, ideas start to grow. Some become algorithms, others become devices that blink and (think?). This section is a garden of those ideas - messy, evolving, and full of surprises.

• EdgeMark
  An automation system for edge-AI tools - which we also use for benchmarking. It helps you generate models, convert them to C/C++ with various optimizations, then compile and test on your device. Project's page
• Cavitation Detection
  Detecting cavitation in centrifugal pumps from vibration signals. We achieved strong performance using SVM and deep learning on a unique dataset curated by Grundfos. Our paper
• Improved Image Transmission
  Improving the bitrate of JSCC image transmission by more than 25%. Reducing the network parameter count by 40% without compromising performance.
• DragonEye
  A platform to collect, store, and display data - built with a gateway board, a server, and an Android app. Project's page
• BuyNow
  Smart store that identifies picked products and deducts their price from the user's credit. Product's page
• Cassava Leaf Disease
  A Kaggle competition to classify cassava leaf diseases. Applied various architectures and techniques to achieve the best performance in the class project competition.
• CNN object detection
  Built MS-VAN model integrating Faster R-CNN, skip-connected autoencoders, and attention modules to predicting bounding boxes on the NWPU VHR-10 dataset. Colab Link
• GAN Adventure
  Exploring generative adversarial networks (GANs) in multiple projects. Music-composer GAN DCGAN image generation CycleGAN
• RL cartpole racing
  Use reinforcement learning to play CartPole. The goal is to run as far as possible without dropping the pole. The agent beat me soundly! Colab Link
• Deep JSCC image transmission
  Reference Paper Implementation Improvement (Decoder in Transmittor) Improvement (Predictor in Transmittor) RNN-based implementation
• Autoencoders Adventure
  Implement AE, VAE, and Disentangled VAE to learn useful features of MNIST images. Colab Link
• And More Stuff
  I've made many more projects, but some weren't lucky enough to make it here :)

RESUME

Download my CV here
last updated: October 2025

MESSAGE ME

I like it when my messaging apps show a red badge on their icons - so feel free to contact me.
That's why I don't answer them :)