Hi, my name is

Ali Abedi.

I am a machine learning engineer.

I build machine-learning-driven software that turns data into real-world solutions. Currently, I'm creating AI-powered, human-centred healthcare products at UHN.

About Me

Hello! I'm Ali, and I love turning messy data into meaningful insights. My journey into machine learning kicked off back in 2010, when I dived headfirst into optical character recognition. Armed with Support Vector Machines and Hidden Markov Models (the cool kids of the time), I built a document image retrieval system that felt like sci-fi made real.

That passion followed me into my Ph.D., where I tackled document image super-resolution using, you guessed it, more machine learning.

Fast forward to today, I’ve had the thrill of building real-world AI-powered systems across banking, healthcare, and beyond. From graphs, images, and videos to sensor signals and natural language, I’ve worked with all kinds of data to craft intelligent, decision-making software that lives out in the wild.

Here are a few technologies I’ve been working with recently:

  • Python
  • Kotlin
  • PyTorch
  • Azure
  • Docker
  • LLM
Headshot

Where I’ve Worked

Scientific Associate @ KITE Research Institute

May 2024 – Present

  • Designed and deployed an ML-driven multimodal sensor platform for remote patient monitoring using smartphones, smartwatches, and sleep sensors. Built feature engineering pipelines and machine learning models on Google Cloud Platform, deployed across multiple hospitals. The models analyzed time-series sensor data to detect social isolation and functional decline, and to predict responsive behaviors.
  • Built a web-based rehabilitation platform with an interactive avatar, connected to device cameras for in-home patient care. Applied deep learning and LLMs with prompt engineering on Azure to assess exercise quality from video-based body joint data, enabling real-time feedback. Models supported regression and binary classification for evaluating movement quality and used natural language generation to deliver feedback.
  • Developed a hybrid LLM and vision transformer framework with prompt engineering for disease classification and localization in chest X-ray images, improving diagnostic accuracy and interpretability.
  • Developed a deep learning framework to analyze facial video data for assessing the affective states of drivers with mild cognitive impairment during autonomous vehicle takeover requests. The system evaluated emotional responses to support safety and behavioral insights.

What I’ve Studied

Postdoc @ University of Toronto

2020 – 2022

  • Focus: Machine learning and Deep Learning for Rehabilitation and Aging.
    I built machine learning, deep learning, and data analysis tools to process data of different modalities from patients and older adults, collected at home, in hospitals, and in long-term care settings, to detect and predict various health outcomes.

Publications

view the archive

What’s Next?

Get In Touch

Although I’m not currently looking for any new opportunities, my inbox is always open. Whether you have a question or just want to say hi, I’ll try my best to get back to you!