AI & Machine Learning
The Real AI Talent Gap and the Sci-Fi Reality of Computer Vision
Is the AI talent gap really about a lack of engineers, or a lack of foresight? Discover how a recent NTU masterclass on Kookree's Few-Shot Learning and real-time computer vision revealed the true challenges of building, deploying, and governing enterprise AI in Singapore.
If you want to see where the future of technology is heading, sometimes you just need to sit in on a Saturday class.
Recently, in my Technopreneur course at NTU, we had Kelvin from Kookree step in as the lecturer to share what his company is building. It sparked a fascinating discussion about how Singapore is racing to catch up in the AI industry and highlighted a massive disconnect in how we are currently building our AI workforce.
The Problem With "AI Talent"
Everyone agrees that hiring AI talent is a challenge, but I think we are misdiagnosing the root cause. The bottleneck isn’t just a lack of engineers; often, the issue lies with traditional hiring managers.
Many leaders don't fully grasp the AI market or the sheer speed at which it is transforming the landscape. They screen for basic technical ability, but identifying true AI talent requires looking deeper.
A real AI professional brings two distinct things to the table:
Foresight: They don't just know what AI does today; they have sharp insights into where the technology is heading tomorrow.
Execution: Their deep technical knowledge is matched by a proven ability to handle real-world application and deployment.
Doing More With Less: Few-Shot Learning
Kelvin demonstrated exactly what that high-level execution looks like through Kookree’s work in Computer Vision. They are utilizing a technique called Few-Shot Learning. Instead of feeding a model millions of data points, this technique allows the AI to hit up to 90% accuracy using only a fraction of the normal training data.
Looking at this from a Go-To-Market (GTM) perspective, the immediate business value is massive, particularly for the defense and security sectors.
However, in the AI industry, speed is the ultimate currency. Having superior technology is only half the battle. To scale rapidly and dominate the security market, leaning heavily into strategic partnership channels will be critical. In this space, you cannot afford to move slowly.
From Text to Tracking in Real-Time
The practical application of Kookree's technology feels like something pulled straight from a sci-fi thriller.
Imagine a security breach. A system flags a suspicious individual, and an operator simply types in a text description: Male, 30s, 1.7m to 1.75m tall, wearing a black shirt and glasses. The AI can immediately scan through live streaming feeds or recorded video to locate and track the suspect based purely on those words. If you have an actual photo of the person? The tracking becomes undeniably precise. For security monitoring and massive data sorting, this completely changes the operational playbook.
The Elephant in the Room
As Kelvin walked us through the capabilities, the energy in the room shifted. The class was undeniably impressed by the sheer power of the technology, but that excitement was immediately followed by the hard, necessary questions: What about ethics? What about privacy?
When a system can track a person through a crowd based on a typed description of their shirt, the line between security and surveillance becomes razor-thin.
The technology is no longer a future concept; it is already here, and it works. The real challenge we face now isn't just about how to build it, but how we go to market, how we deploy it, and how we govern it responsibly.