Jean-Baptiste Kempf, best known as the lead developer behind the VLC Media Player, is now applying his experience in open-source video infrastructure to a different computing frontier: real-time control systems for robots and autonomous devices.
Kempf is the founder of Kyber, a Paris-based startup developing an infrastructure layer designed to synchronize video, audio, sensor data, and control inputs with minimal latency. The system is positioned for use cases where operators and machines are physically separated from the underlying compute and execution environment.
The company has raised a $5 million funding round led by venture capital firm Lightspeed, which has also backed companies including Anthropic and Mistral AI. Lightspeed framed its investment in the context of “physical AI,” stating that such systems depend heavily on robust underlying infrastructure.
Kyber’s core product is an SDK that focuses on real-time coordination between remote operators and distributed devices. Kempf describes the system as addressing environments “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.”
The startup’s technical approach draws heavily from streaming architecture. Kempf previously worked on cloud gaming infrastructure at Shadow, and Kyber’s design reflects that background in low-latency video transmission combined with IoT device optimization.
From Video Streaming to Physical AI Infrastructure
The company’s origins are rooted in video streaming systems, a lineage that traces back to Kempf’s work on VLC and later cloud gaming platforms. Kyber extends those principles into physical systems where latency becomes operationally critical.
Kempf has emphasized that in real-world machine control, timing constraints are strict. “If you control things in the real world, every millisecond matters,” he said, underscoring the importance of minimizing lag between operator input and machine response.
The company argues that while similar systems exist in niche deployments such as remote driving, most are limited in scale. Kempf noted that existing implementations often manage only a few thousand vehicles, whereas future deployments may require coordination across millions of devices.
Scaling Remote Operations and Device Observability
Kyber’s architecture also targets a challenge that becomes more complex at scale: observability. As fleets of autonomous systems grow, ensuring that devices are functioning correctly and responding as expected becomes a core engineering requirement.
The company positions its system as applicable to a range of scenarios, from remote software updates to full operational control of distributed hardware fleets. This includes robotics, drones, and remote IT infrastructure management.
Kyber is also exploring the remote IT access space, a market often associated with enterprise tools such as Citrix. Kempf has indicated that Kyber aims to extend beyond traditional remote desktop models into broader infrastructure control systems.
Open Source Foundations and Enterprise Deployment
Kyber is built on a hybrid model. Its core software remains open source, reflecting Kempf’s long-standing involvement in community-driven development. Alongside this, the company offers enterprise-grade deployments and services, including forward-deployed engineering support.
Forward-deployed engineers form a significant portion of Kyber’s approximately 25-person team. The startup operates from Paris, with additional offices in San Francisco and Singapore, reflecting its target of global enterprise deployment.
According to the company, Kyber is already in commercial use across sectors including defense, telecommunications, robotics, and artificial intelligence systems.
Market Position and Technical Direction
Kyber’s strategic focus is concentrated on three main areas: robotics, drone systems, and remote IT access. Each reflects different aspects of distributed machine control, unified by the requirement for low-latency synchronization between operators and devices.
The company’s underlying premise is that future AI systems will increasingly interact with physical environments, making real-time infrastructure as critical as model performance. Lightspeed described this dependency in its investment announcement, noting that “physical AI is only as good as the underlying systems running it.”
While still early in scale, Kyber’s positioning reflects a broader industry shift toward infrastructure designed for embodied AI systems rather than purely digital applications.
