My full name is Muhammad Ammar Ul Hassan, but you can call me Ammar. I’m a computer vision researcher and engineer based in Galway, Ireland. I did my PhD at Soongsil University in Seoul, spent time working on real-world AI systems and their deployment at DeltaX.ai, and I’m currently a Postdoctoral Researcher at the University of Galway. Feel free to reach out if anything here overlaps with what you’re working on.
A bit more about me
I’m originally from the beautiful city of Islamabad, Pakistan. I spent nine years living in Seoul, South Korea, which was a great experience on many levels. I moved to Galway, Ireland in late 2025 and have been enjoying the city and settling into its warm and welcoming community.
I’m a proud Manchester United supporter and have been following the club since the Sir Alex Ferguson years. Those were special times and I’m still very much hoping for more titles ahead, fingers crossed.
What I do
My work has covered generative models, driver monitoring, 3D scene understanding, and edge deployment. Different problems, but with the same underlying interest: vision systems that work reliably in practice.
Right now I’m part of the ATHENA project, focusing on detecting impaired drivers using thermal cameras. The project research also explores Vision-Language Models (VLMs) for driver monitoring systems, specifically benchmarking current VLMs and evaluating existing VLM assessment metrics in this domain. I’m also working on open-vocabulary detection frameworks for in-cabin and outdoor scenes, where you can’t always predict in advance what you’ll need to detect.
Before that
At DeltaX.ai in Seoul, I worked on AI problems related to automobiles and smart factories.
On the automotive side, the work included in-cabin monitoring (seatbelt detection, drowsiness, occupancy from IR cameras), a road scene segmentation pipeline for ADAS, and deploying the resulting models on Jetson and TI edge hardware. A significant part of this involved getting models that performed well in training to run efficiently on constrained hardware without losing accuracy on what mattered.
For smart factories, I worked on a 27-camera vision system covering human intrusion detection and surface inspection, including crack and hole detection.
My PhD focused on image synthesis using GANs, covering domain translation, style transfer, and facial attribute transfer using supervised, semi-supervised, and unsupervised learning techniques. I designed various model architectures and training strategies, with font synthesis as the primary application domain. The problem of separating what a character is from how it is drawn turned out to be a challenging and genuinely interesting problem for unsupervised representation learning.
Sensors and tools I work with
I’ve worked with RGB, fisheye, depth, and IR/thermal cameras across different projects, along with LiDAR, IMU, and CAN bus data. I mostly use PyTorch as a deep learning framework, though I’ve also used TensorFlow in some of my PhD paper implementations (SKFont, SkelGAN). I have deployed models through ONNX, TensorRT, and DeepStream onto Jetson, TI, and Hailo hardware.
Open to collaboration on
Driver monitoring and in-cabin perception, VLM and LLM evaluation for real-world vision tasks, diffusion models and domain adaptation, current generative approaches, and efficient edge AI deployment. If your work touches any of these, feel free to reach out.
Education
- Ph.D. Computer Science & Engineering, Soongsil University, Seoul (2018–2023), advised by Prof. Jaeyoung Choi
- M.S. Computer Science & Engineering, Soongsil University (2016–2018) · International Graduate Research Scholar
- B.S. Software Engineering, International Islamic University, Pakistan (2009–2013) · Federal Government Scholar
Get in touch
Email: ammar.instantsoft@gmail.com · GitHub · LinkedIn · Google Scholar