Mumbai man develops Bell’s palsy after returning from Singapore, builds AI face-tracking app for facial rehabilitation

16 hours ago 1
ARTICLE AD BOX

A Senior Developer Advocate from Mumbai, who was on an official trip and “having the best time of his life”, was left in shock after returning home to discover that the right side of his face had stopped working. Ali Mustafa shared his ordeal on X formerly Twitter, detailing how he had been attending conferences and speaking to people about artificial intelligence — something Mustafa described as the “core part of his job” and he “loves” doing.

Mustafa, who has been involved in developer advocacy and public speaking for the past 10 years, said his life took an unfortunate turn after returning to Mumbai from Singapore.

“I had to go to the emergency because my right side was starting to fail, especially my face. It was not responding, and I was not able to blink. I was not able to do whatever I was able to do on the left side,” he said.

He stated that he had developed Bell’s palsy — a condition that usually occurs in people who suffer from infections due to extreme cold. However, he said he could not comprehend the reason behind the condition, as both Singapore and Mumbai are humid places.

“So it was very unexpected, but it was detected, and it happened. Overnight, I was not able to move the right side of my face. As much as demotivating that was, not able to speak, not being able to do what I love doing was more painful,” he said.

AI tool designed for facial rehabilitation

Mustafa later built an AI-powered tool called “Mirror” to track facial recovery exercises after being diagnosed with Bell’s palsy. He shared the project online, explaining how the platform uses facial landmark detection and symmetry analysis to monitor rehabilitation progress and generate reports for physiotherapists.

“But I'm a developer at heart and I love working around machine learning and AI models,” he said while introducing the project. He added that he had previously created courses that taught “more than 500,000 students” how to get started in machine learning.

The developer explained that the condition usually lasts for weeks and that recovery depends heavily on regular exercise. Wanting a better way to monitor improvement, he decided to combine his technical background with his recovery process.

“So I gathered some courage, sat down with my laptop, booted up a lot of AI models and created Mirror,” he said.

According to him, the platform helps people with conditions such as Bell’s palsy perform facial exercises while evaluating facial symmetry between the left and right sides of the face.

“It helps you to do exercises and also evaluate them based on the symmetry on the left side of your face to the right side of the face,” he explained.

Tracks progress and generates reports

The tool allows users to record exercises daily and maintain a journal documenting their recovery journey. It also creates detailed reports that can be reviewed by physiotherapists to provide feedback and assess progress.

The developer demonstrated how users can revisit individual exercises, track exercise duration and even watch time-lapse recordings showing facial improvement over time.

“The idea is to not just do the exercises, but also see the progress,” he said.

Uses machine learning and facial landmarks

The backend system is powered by machine learning models and uses MediaPipe facial landmark technology. According to the creator, the model first captures a neutral facial expression before dividing the face into two sections for comparison.

The system records more than 400 facial points and analyses symmetry across different exercises such as eyebrow movement and smiling.

“For every exercise, it will do that. And at the end, it will normalize the score and then show it to my physiotherapist,” he explained.

The tool reportedly relies on nearly 478 3D facial benchmarks and landmarks, including eyebrow, eyelid, lip and eyeball tracking.

Mobile-first approach

The developer said the project was designed primarily for mobile devices so users could easily hold up their phones like a mirror while performing exercises.

“The model which we are using here can run on any browser and also run on mobile device. This is a mobile-first approach,” he said.

He also revealed plans to deploy the project publicly and possibly upload the code to GitHub so others with similar conditions can benefit from it.

“If you know someone who has this condition and who would benefit from this, I’ll create a GitHub repository and also create a sample deployment,” he added.

Read Entire Article