From 6 Billion Downloads to Robots on the Street
If you have ever played a video file on a computer, there is a good chance you have used VLC Media Player — the free, open source video player recognisable by its orange traffic-cone icon. With more than 6 billion downloads to its name, VLC is one of the most successful pieces of open source software ever released. Now its lead developer, French engineer and entrepreneur Jean-Baptiste Kempf, has turned his attention to an entirely different technical challenge: building the foundational open source robot infrastructure that will allow humans — and eventually AI agents — to control robots, drones, and remote devices in real time, at global scale.
The startup he founded, Kyber, has just raised a $5 million seed round led by Lightspeed Venture Partners — a firm that has also backed AI companies Anthropic and Mistral AI. The investment reflects a broader conviction in the technology industry that the era of physical AI, where intelligent systems operate in the real world rather than purely in software, is arriving faster than most people anticipated. According to TechCrunch, Kempf believes "hundreds of millions of robots and drones" will be moving through public spaces within a few years — and that the software layer controlling them does not yet exist in any robust, open, or widely accessible form.
For developers, IT architects, and technology decision-makers tracking the intersection of European open source innovation and emerging AI infrastructure, Kyber represents something unusual: a technically credible, open-source-first approach to a problem that every major technology company involved in robotics, autonomous vehicles, and remote operations will eventually need to solve.
What Kyber Actually Does — and Why Latency Is Everything
Kyber's core product is a software development kit (SDK) that synchronises video feeds, audio streams, sensor data, and control inputs with minimal latency. The startup's name is a nod to the lightsaber crystals in Star Wars — a deliberate signal of the company's obsession with speed. "If you control things in the real world, every millisecond matters," Kempf has said. In practical terms, this means Kyber is solving the same class of problem that makes remote surgery, autonomous vehicle teleoperation, and drone delivery commercially viable: ensuring that the person or system issuing a command and the machine executing it are never meaningfully out of sync.
The technical framing Kempf uses is precise and worth understanding: Kyber is built for "all the use cases where the person who's operating is not in the same place as the compute, which is not in the same place as the action." This is the defining architecture of physical AI — distributed control across geography, with real-time constraints that traditional enterprise software was never designed to meet. Research from McKinsey's technology trends reports has consistently highlighted low-latency connectivity and edge computing as foundational requirements for the next generation of industrial automation.
Kyber's approach to solving latency is rooted in video-streaming technology — a natural inheritance from Kempf's work on VLC. The startup actually began as a side project he built while serving as CTO at Shadow, a cloud gaming company, where low-latency video delivery was also a core engineering constraint. The technical lineage is direct: the same principles that make a video stream feel instantaneous are being applied to make a robot respond instantly to a remote operator's inputs.
"Physical AI is only as good as the underlying systems running it."
— Lightspeed Venture Partners, announcing their investment in KyberBut Kyber is not simply repurposing streaming software. The other half of its optimisation work is IoT-specific: tuning performance to the available compute on any given device, whether that is a high-powered industrial robot or a constrained sensor node operating at the edge of a network. This combination — streaming expertise plus IoT performance tuning — is what Kyber argues makes its infrastructure genuinely novel rather than a reconfiguration of existing tools.
An Open Source Model With Enterprise Teeth
For developers and IT decision-makers familiar with the open source software ecosystem, Kyber's business model will feel immediately recognisable. True to Kempf's roots — and consistent with how projects like Linux, Kubernetes, and HashiCorp's Terraform originally scaled — Kyber's core software is open source, designed to be used by the widest possible community. The company monetises through a productised enterprise version, targeting organisations that need commercial support, compliance guarantees, and the kind of reliability SLAs that production deployments demand.
This model matters for European audiences in particular. The region's technology policy environment, shaped by the European Union's emphasis on digital sovereignty, open standards, and reducing dependency on proprietary American or Chinese platforms, has made open source infrastructure a priority for both governments and large enterprises. Initiatives like the European Commission's open source strategy and the continued expansion of GDPR compliance requirements mean that organisations increasingly want to understand and audit the software controlling their systems — something that open source infrastructure makes structurally possible in a way that proprietary solutions do not.
Kyber also offers something less common in pure software startups: forward-deployed engineers (FDEs). These are specialists embedded directly with customers to handle custom deployment and integration work — a model more associated with defence contractors and data infrastructure companies like Palantir than with typical software-as-a-service businesses. FDEs make up a significant portion of Kyber's current 25-person team, which is headquartered in Paris with offices in San Francisco and Singapore.
Target Sectors: Robotics, Drones, and the Unglamorous Opportunity in Remote IT Access
Kyber has identified three initial segments to focus its commercial efforts: robotics, drones of all types, and remote IT access. The first two are intuitive given the physical AI framing, and they represent markets where the absence of standardised, open infrastructure is already a visible bottleneck. As Wired's robotics coverage has documented, many of the largest operators in autonomous vehicles and drone logistics have built their own proprietary remote-control stacks — expensive, bespoke systems that took years and tens of millions of dollars to develop and that they have no commercial incentive to share.
Kempf's argument is essentially a classic open source disruption case: the problem is generic enough that building a shared, openly available solution benefits the entire industry, while proprietary solutions create redundant engineering effort and fragmentation. "The companies that tried to solve it spent years and tens of millions building custom solutions they'll never share. We're building the version everyone else can use," reads Kyber's own careers page — an unusually candid statement of competitive positioning.
The scale argument is particularly compelling for enterprise IT decision-makers. Kempf notes that the largest existing fleets managed with custom remote-control software today number around 2,000 to 3,000 vehicles. The infrastructure requirements for managing millions of devices simultaneously are qualitatively different — not just bigger, but architecturally different in terms of observability, fault tolerance, and update management. Being able to push software updates to a fleet of devices remotely, without physically reaching each one, is a capability that matters even at modest scale today. At the scale Kyber is designing for, it becomes a baseline requirement.
The third segment — remote IT access — is less dramatic but commercially significant. Kyber positions itself as an alternative to established players like Citrix in the remote access market, though Kempf frames the ambition as broader than that comparison suggests. The remote IT access market is substantial, and demand for Kyber's solutions in this segment has reportedly been particularly strong during its early commercial phase.
Why Observability Becomes Critical When AI Agents Run the Fleet
One of the less-discussed but technically important dimensions of Kyber's offering is observability — the ability to know, in real time, whether distributed systems are functioning correctly. For IT professionals, observability is already a first-class concern in cloud-native architectures, where tools like Datadog, Prometheus, and OpenTelemetry have become standard infrastructure. As Gartner has noted in its infrastructure research, observability requirements become significantly more complex as systems scale and as human oversight is replaced or augmented by automated decision-making.
Kyber's infrastructure is being built with this transition in mind. Today, a human operator monitoring a remote drone or robot can notice when something looks wrong and intervene. In the near-term future Kempf is designing for, AI agents — not people — will be managing entire fleets and networks. The observability layer that lets those agents know whether their commands are being executed, whether sensor data is reliable, and whether a device has failed becomes a safety-critical system component, not just an operational convenience.
| Kyber Target Segment | Core Use Case | Key Technical Requirement | Current Deployment Status |
|---|---|---|---|
| Robotics | Remote fleet control and teleoperation | Sub-millisecond latency, sensor sync | Commercial deployment active |
| Drones | Autonomous and semi-autonomous UAV management | Real-time video + control synchronisation | Commercial deployment active |
| Remote IT Access | Secure remote device and infrastructure management | Low-latency access, scalable architecture | High demand, commercially active |
| Telco | Network device monitoring and remote operations |
