Kexxu's Smart Cameras Are Already Watching Steel Mills and Tracking Eyes

The Amsterdam-based AI and robotics company is building a portfolio of specialized computer vision products, from wearable eye trackers to industrial quality control systems.

About Kexxu

Published

If you ask a steel factory manager what they need, a computer vision startup from Amsterdam probably isn't the first answer. But Kexxu, a company of fewer than ten people, is betting that its custom-built smart cameras are the answer to a dozen different questions across industrial and research settings. Their strategy is methodical, if scattered: build the hardware, train the model, and sell the entire machine perception stack for a specific, often overlooked, workflow. It is a wedge strategy that trades scale for precision, and its success hinges on whether a small team can master enough verticals to build a real business.

A Portfolio of Specialized Eyes

Kexxu does not sell a single product. Instead, it operates as a project shop for applied computer vision, developing turnkey systems that combine custom or adapted camera hardware with proprietary deep-learning models [LinkedIn]. The company's public portfolio reads like a list of disparate pilot projects. There are smart cameras for monitoring steel production and checking the quality of reusable cups [LinkedIn]. There is OpenEye, a wearable gaze-recording device, and a line of 360-degree thermal cameras for anonymous marketing analytics [Dealroom, July 2025]. Other listed offerings include AI tools for documentary transcription and biometric data collection for psychological research [LinkedIn]. The common thread is a focus on complete, often edge-deployed solutions where the AI processing happens on the device itself. One product, the Kexxu Deep Learning Unit, is described as a GPU-accelerated edge device built for harsh industrial environments [Kexxu].

The Hardware-Integrated Wedge

The company's apparent moat is integration. In a market flooded with generic AI APIs and off-the-shelf sensors, Kexxu is betting that customers in niche industrial and research fields will pay for a system that works out of the box. The value proposition is reducing the engineering burden. For a factory, this might mean a camera unit that performs quality inspections locally without sending sensitive video data to the cloud. For a researcher, it could be a pair of eye-tracking glasses with a six-hour battery life and a companion app for uploading recordings to free cloud analysis tools [Kexxu]. A 2024 post by an associate highlighted a key design choice: Kexxu's 'Industrial AI camera processing unit' integrates the compute directly, aiming to replace external and expensive separate units [Mees van der Hart - Noord Loom | LinkedIn, 2024]. This focus on the edge aligns with a projected market for smart cameras with neural processing units (NPUs) that could reach $76.79 billion by 2035 [SNS Insider].

Early Traction and Team Build

Evidence of commercial traction is limited but points to active deployment. One source states that Kexxu has created software for over 100 AI cameras running in steel factories globally, with plans to add more [Valerio Scacco - Università di Roma Tor Vergata | LinkedIn]. The company is estimated to have between two and ten employees [Dealroom, July 2025]. A 2024 post noted the company was moving to a new office more than twice the size of its previous one, citing rapid growth [Laurien Schreuder - Snowapple Collective | LinkedIn, 2024]. The technical team appears lean but focused. Co-founder Tijmen De Vries has led the company since October 2020 [LinkedIn]. Jurriaan Schreuder is listed as Chief Technology Officer, with a background in C++ and systems engineering [The Org]. Another engineer, Jaap Oosterbroek, has worked on building real-time eye-tracking neural networks for Raspberry Pi platforms at Kexxu [LinkedIn].

Metric Value
Smart Cameras with NPU Market 76.79 B USD by 2035
Kexxu Eye Helmet Fit Price 6194.7 USD
Reported AI Camera Deployments 100 units

The Fragmentation Risk

The biggest challenge for Kexxu is focus. A portfolio spanning steel mills, cup manufacturers, documentary filmmakers, and sports analysts suggests a consultancy model, not a scalable product company. Each vertical requires deep domain knowledge, custom model training, and likely a unique sales motion. The company has not publicly disclosed any funding rounds, which raises questions about its runway to develop and sell into these varied markets simultaneously [Dealroom, July 2025]. Without a clear lead investor or a disclosed war chest, the path to achieving density in any one sector is steep. The competitive set is also fragmented but real. In industrial vision, giants like Cognex and Keyence dominate, while specialized startups like Kornea.AI attack similar niches. For wearable eye tracking, established research equipment companies like Tobii command premium prices, against which Kexxu positions its OpenEye device as an affordable alternative [Kexxu].

The Next Twelve Months

For Kexxu, the coming year will be about proving repeatability. The signal to watch is whether the company begins to coalesce around one or two primary verticals where it can demonstrate not just a working camera, but a growing list of paid deployments and renewal contracts. The mention of over 100 cameras in steel factories is a start, but the next step is converting those deployments into a predictable revenue stream and a referenceable customer base. Another milestone would be a disclosed funding round, which would provide both capital and a validation point for its integrated hardware-plus-AI thesis. The company's move to a larger office suggests internal confidence, but the external proof will be in landed enterprise pilots and a clearer public narrative.

The ideal customer profile here is a mid-size industrial operator or a specialized research lab. They have a clear, repetitive visual inspection task or data collection need but lack the in-house AI expertise to build a solution from generic components. They are pragmatic buyers looking for a fixed-cost system that works in their specific environment, not a platform to build upon. For them, Kexxu's bet on integrated edge devices is the entire value proposition.

Realistically, Kexxu competes in two separate arenas. In industrial automation, it faces entrenched incumbents and must compete on customization and price. In the research and analytics tool space, it competes on affordability and portability against high-end, institutional-grade equipment. Its path is to out-specialize the generalists and underprice the specialists, a difficult but not impossible lane if execution is flawless.

Sources

  1. [LinkedIn] Kexxu company profile | https://www.linkedin.com/company/kexxu
  2. [Dealroom, July 2025] Kexxu company information | https://app.dealroom.co/companies/kexxu
  3. [Kexxu] Kexxu product pages | https://kexxu.com/
  4. [Mees van der Hart - Noord Loom | LinkedIn, 2024] Post on industrial AI camera unit | https://www.linkedin.com/posts/jurriaan-schreuder-968b7731_new-product-page-in-the-making-at-kexxu-activity-7198319308450480129-Iyet
  5. [SNS Insider] Smart Camera with NPU Market report | https://www.snsinsider.com/
  6. [Valerio Scacco - Università di Roma Tor Vergata | LinkedIn] Post on Kexxu camera deployments | https://www.linkedin.com/in/valerio-scacco
  7. [The Org] Kexxu organization chart | https://theorg.com/org/kexxu
  8. [Laurien Schreuder - Snowapple Collective | LinkedIn, 2024] Post on Kexxu office move | https://www.linkedin.com/in/laurien-schreuder

Read on Startuply.vc