The Company That Started Before Anyone Was Watching

In 2013, before most people had heard the words machine learning in a commercial context, a founder in Egypt started a company called QEYE.

It was a quality inspection system for the textile industry, built on classical computer vision. Not neural networks. Not deep learning. Classical algorithms analysing images to detect defects in fabric.

The kind of work that in 2013 required painstaking manual calibration for every new application and had almost no commercial infrastructure around it.

Four years later, he started DevisionX.

The technology had shifted. Neural networks and deep learning were replacing classical computer vision. GPUs were becoming commercially accessible.

The market was beginning to understand that machines could see in ways that had real industrial value.

He has been building in this space ever since. Ten years. Three distinct technology eras. Multiple business model pivots.

And a company that is still standing, still serving clients, and now positioning for a fundraise at the precise moment the rest of the world has caught up to what he has been doing all along.

The timing is not coincidental.

---

The Core Problem

Every large institution in the world right now is trying to figure out how to apply AI to its processes.

The founder describes this with a directness worth quoting:

Whether they speak Arabic or English, whether they are in Egypt, Saudi Arabia, or the UK, they are all in the same place.

They know they need AI. Although some do not know how to start.

They do not know what their first use case should be. They do not know whether their data is ready, whether their teams are ready, or how to measure what success would look like.

That observation, that enterprise AI adoption is still stuck at assessment rather than implementation across geographies, cultures, and languages simultaneously, is a significant ecosystem intelligence finding.

The AI capability exists. The institutional readiness does not.

For regulated sectors specifically, the problem has an additional layer. Government agencies, banks, healthcare systems, and insurers operate under data sovereignty requirements that prohibit them from sending sensitive information to public cloud APIs.

They cannot simply plug into a foundation model from a foreign technology company and process their citizens’ data through it.

They need AI that lives inside their own infrastructure, under their own control, subject to their own jurisdiction.

That is not a niche requirement. It is a structural condition for an enormous portion of global enterprise spending.

And it is a condition that most AI companies, built for the public cloud market, are structurally unable to serve.

---

The Strategic Decision Layer

DevisionX has made two distinct strategic pivots across its ten-year history and is currently in the middle of a third.

The current pivot, from computer vision alone to multimodal AI combining vision and text, reflects a genuine technology threshold.

When large language models demonstrated the ability to reason about images in 2024, the competitive landscape for pure computer vision companies changed fundamentally.

A system that could only see was now competing against systems that could see and think.

DevisionX’s response was not to abandon its computer vision expertise but to extend it.

The see-think-act architecture is the current expression of that extension.

See – is the computer vision layer, detecting and interpreting what is in an image or document.

Think – is the multimodal reasoning layer, adding context, language, and analysis to what was seen.

Act – is the output layer, the decision or physical action that follows from the combined perception and reasoning.

For regulated sector clients, Act manifests as a structured decision recommendation that a human approves before implementation.

For physical AI and robotics, Act is increasingly literal.

The behaviour learning dimension is the most forward-looking element of the current strategy and worth examining carefully.

The founder describes a shift in how robots and physical AI systems are trained.

Previously, training meant feeding images and videos to a model and teaching it to recognise objects.

Now training means showing a system human behaviour, a person cutting an apple, a technician fixing an engine, and letting the system learn the task from watching it performed.

The creation of these behaviour videos is itself a new category of work, a new type of job that did not exist five years ago.

DevisionX is positioning to work in this space precisely because it has spent a decade building the technical infrastructure that makes using this data meaningful.

---

Ecosystem Context

DevisionX started in Egypt in 2017 and opened a London office in 2023.

Egypt first was not a limitation of ambition. Rather, it was where the founding team was and where the early clients were.

The London office came six years later, specifically to access the European and UK markets with a presence that those markets recognise and trust.

DevisionX currently navigates GDPR for European clients, data protection requirements in the MENA (Middle East and North Africa) region, all simultaneously.

Navigating multiple regulatory environments from early stage produces a compliance fluency that single-market companies develop much later, if at all.

The AI readiness assessment service the company now offers as a paid consulting layer also reflects an ecosystem observation worth noting for investors.

---

Observed Patterns

The founding history across two companies before DevisionX reached its current form is the most important credibility signal in this record.

Each venture added a layer of domain knowledge and market understanding that the next one was built on.

The Digified experience in digital identity is particularly relevant. Identity verification is one of the most technically demanding computer vision applications because the consequences of error are severe and the regulatory requirements are strict.

That experience is directly relevant to the regulated sector focus DevisionX has now adopted.

The company’s survival across the transition from classical computer vision to deep learning to generative AI and now to multimodal and physical AI is itself an analytical signal.

Most companies built on a single technology generation do not survive the transition to the next one.

DevisionX has navigated three such transitions in ten years. Each one required the team to candidly assess whether their existing capabilities were still competitive and to add the new layer before they were forced to.

The decision to fundraise only after commercial proof across regulated sectors, targeting Q1 2027, reflects a founder who has bootstrapped long enough to understand the cost of raising capital before the product thesis is sufficiently proven.

That patience, holding the fundraise until the commercialisation story is clear, is a more disciplined capital strategy than many better-funded companies demonstrate.

---

Open Variables

The physical AI and behaviour learning direction, while technically coherent, represents a significant expansion of the company’s scope at a moment when it is also attempting to productise its multimodal AI offering for regulated sectors.

Managing two parallel technical directions, one serving enterprise software clients and one building toward robotics and physical AI infrastructure could be an operational stretch that the current narrative acknowledges without yet fully resolving.

The competitive positioning at the intersection of computer vision and generative AI is credible as a statement of current capability.

How durable that positioning is as larger, better-resourced companies increasingly combine the same capabilities is a question that depends on how quickly DevisionX can convert its technical edge into contracted, referenceable enterprise deployments.

The window in which the combined expertise advantage is most distinct is the current period.

---

Why This Matters

For founders building in AI infrastructure for regulated sectors, this case makes a specific argument about the commercial advantage of building in markets where trust must be earned before any transaction can happen.

The sovereign AI requirement, the insistence that regulated sector clients cannot use public cloud APIs, is a feature of the market that most generalist AI companies treat as a barrier.

DevisionX has oriented its entire technical architecture around it. That orientation, however difficult to build, creates a defensible position in a market segment that the largest AI companies are structurally reluctant to serve.

For investors evaluating AI companies from Egypt and the broader MENA region, the DevisionX case surfaces a founder whose technical depth was accumulated across a decade of building before the current AI wave made the category investable.

The risk profile here is different from a first-time AI founder riding the generative AI wave. It is a ten-year operator who has survived multiple technology transitions and is now positioned at the intersection of two converging trends, multimodal AI and sovereign infrastructure, at the moment both are becoming enterprise purchasing priorities.

For DFIs and development organisations, the AI readiness gap the founder describes across medium and large enterprises in Egypt and the Middle East is an ecosystem observation that deserves dedicated attention.

If every large institution in the region is at the same early stage of AI adoption, stuck at assessment rather than implementation, the bottleneck is not capability.

It is trusted guidance at the point where institutions are making first decisions.

Supporting companies that provide credible AI readiness assessment, and implementation guidance creates multiplier effects across the enterprise AI adoption curve.

---

Final Strategic Takeaway

The founder’s advice to upcoming founders is one sentence.

Find your differentiation and change fast, monthly or quarterly, not yearly.

That advice is worth sitting with in the context of his own history.

The fundraise may not yet be opened. The largest client wins may not yet be public. The physical AI chapter is still being written.

But the founder who started building before the wave arrived is now, for the first time in ten years, building at precisely the moment the market is looking for exactly what he has been building all along.

The timing finally matches the direction.

---

This article is drawn from an in-depth founder interview conducted by Afriq IQ with Mahmoud AbdelAziz, founder of DevisionX. Selected insights and observations are published here.