Designing AI That Matters: My Research Journey at the University of Waterloo


When people hear “Artificial Intelligence,” they often picture breakthrough algorithms or futuristic robots. In reality, designing AI that truly matters means building systems that solve real human problems, and doing it with purpose, ethics, and context in mind.

My six-month research internship at the University of Waterloo’s Ubiquitous Health Technology Lab (UbiLab) was a deep dive into that reality. It wasn’t about chasing the latest AI trend, it was about engineering intelligence that works in the real world, under real constraints, for people who actually need it.

AI in Service of a Bigger Mission

The project aimed to improve independent living for seniors and people with health challenges. We developed an ambient activity recognition and behavioral anomaly detection system that could operate without cameras or microphones : protecting privacy while still providing actionable insights.

This required:

  • IoT Sensor Integration: Zigbee-based motion, door, and appliance sensors connected via MQTT, Domoticz, and InfluxDB.

  • Feature Engineering: Transforming millions of raw time-series readings into behavioral patterns.

  • Machine Learning Models: Unsupervised clustering and anomaly detection for daily activities like cooking, showering, or sleeping.

  • Interpretability: Models that healthcare professionals could actually understand and trust.

The AI was central, but it was never the whole solution. It lived inside a carefully designed, ethically grounded system built for long-term use.

Designing AI Is More Than Just Writing Models

In textbooks, AI might seem like a problem of model selection and parameter tuning. In reality, designing AI that matters means:

  • Understanding the domain and its constraints.

  • Deciding what to measure and how to transform that data into useful features.

  • Choosing algorithms not only for performance but for explainability and adaptability.

  • Integrating AI into systems that run reliably and securely in real environments.

In other words, the model is just one piece of the architecture. The real value is in orchestrating all the parts into a system that delivers real impact.


AI as a Productivity Multiplier, Not a Replacement

A key takeaway from this project is that AI is not here to replace humans, it’s here to amplify them.
Our anomaly detection models automated what would otherwise take hours of manual log inspection. That freed me to focus on problem-solving at a higher level: improving system generalization, ensuring privacy compliance, and thinking about how this could scale to more homes.

If AI can replace a repetitive task, that’s a win, because it pushes humans toward creativity, strategy, and innovation.


A Shift in Perspective

Before Waterloo, I saw AI mostly as an algorithmic challenge. After six months in research, my perspective has evolved:

  • AI is a strategic component within a larger engineered system.

  • Its impact depends on context, integration, and human oversight.

  • The best AI is not necessarily the most complex, it’s the one that fits the problem, the environment, and the people using it.

This is what designing AI that matters really means: creating intelligence that is useful, ethical, and sustainable. It’s a lesson I’ll carry forward into every AI project, whether in research, entrepreneurship, or industry.


— Yazan El Mahmoud | LinkedIn

Comments

Post a Comment

Popular posts from this blog

Engineering Studies in Europe vs North America: What Every Student Should Know

Image
Photo by  ESCP When it comes to studying engineering, both Europe and North America offer incredible opportunities — but they do so in very different ways. I’ve had the chance to study engineering in France at the UTC_Sorbonne University , and soon I’ll be heading to the Georgia Institute of Technology in the U.S. for my Master’s in Computer Science with a specialization in Artificial Intelligence. Here’s what I’ve learned from navigating both worlds. Program Structure: Theory vs Application One of the first differences you’ll notice is how theoretical French engineering education can be, especially in the first years. At UTC, and in many French engineering schools, the first two years are a common core — no matter your specialization, you study advanced mathematics, physics, chemistry, mechanics, and even computer science. The idea is to form a “generalist” engineer before you dive deep into your field. And the mathematics? Let’s just say this: you don’t just use theorems — ...
Read more

From France to the “MIT of Canada”: Why I Chose a 6-Month Research Internship at the University of Waterloo Over a Company Placement

Image
  Courtesy of the University of Waterloo. When most engineering students approach their final years of study, they often set their sights on internships in established companies, the kind that promise industry exposure, networking opportunities, and the comfort of a corporate structure. I chose a different path. Instead of joining a company, I packed my bags, flew across the Atlantic, and began a six-month research internship at the University of Waterloo,  often called the MIT of Canada . Why? The reasons were as personal as they were professional. Aligning Research With My Entrepreneurial Vision Before coming to Waterloo, I had already been working on entrepreneurial projects in France related to smart health and assistive living,  from fall detection systems for seniors to IoT-based health monitoring solutions. When I discovered the Ubiquitous Health Technology Lab (UbiLab) at Waterloo, I knew it was the perfect fit. The lab’s research on ambient sensor networks...
Read more

When Code Becomes Secondary and Logic Reigns

Image
Source  The history of programming is the history of translation. For over half a century, humans have been painstakingly converting ideas into languages that machines can execute. From the rigid world of assembly to the elegance of Python, every leap forward in programming has been about one thing: narrowing the gap between human thought and machine action . But the arrival of Large Language Models (LLMs) has brought us to a turning point. The value of an engineer will no longer be measured by how quickly they can write code, but by how precisely they can design logic. Large Language Models have shifted the center of gravity from syntax to reasoning, from typing instructions to architecting systems.  The future belongs to those who can think in pure structure and let AI handle the translation into machine language.  I didn’t arrive at this conclusion by reading headlines,  I arrived at it by building real AI systems, watching code become the least important part o...
Read more