The Man Behind VLC Is Now Building the Backbone of Physical AI
Jean-Baptiste Kempf is a name that may not ring a bell for the average internet user, but his work has touched nearly every person on the planet. As the lead developer of VLC Media Player, Kempf helped create one of the most ubiquitous pieces of open-source software in history, with over 6 billion downloads to date. Now, he’s turning his attention to a challenge that could be even more transformative: building the infrastructure that will power millions of robots and drones in the real world.
His new startup, Kyber, is developing an SDK (Software Development Kit) designed to synchronize video, audio, sensor data, and control inputs for remote devices with minimal latency. This isn’t just another incremental improvement in robotics—it’s a foundational layer that could enable physical AI to operate at scale, from autonomous delivery drones to teleoperated surgical robots.
But how did a video-streaming expert end up in robotics? And why did Lightspeed Venture Partners, the firm behind investments in Anthropic and Mistral AI, bet $5 million on this Paris-based startup? Let’s break down the technology, the vision, and the implications of Kyber’s work.
Why Robots Need a “VLC for Real-Time Control”
The Latency Problem in Robotics
Robots and drones are only as useful as their ability to respond in real time. Whether it’s a self-driving car avoiding an obstacle or a drone delivering medical supplies, every millisecond counts. Traditional cloud-based control systems introduce lag, which can be catastrophic in high-stakes scenarios.
Kyber’s solution? Treat robot control like video streaming—but with even stricter performance requirements.
- Video streaming (VLC’s domain): Optimized for smooth playback, even with network fluctuations.
- Robot control (Kyber’s domain): Requires sub-100ms latency for safe, precise operation.
Kempf’s background in cloud gaming (via his CTO role at Shadow) gave him firsthand experience in low-latency remote control, a skill set that directly translates to robotics.
The Three Pillars of Kyber’s Tech
Kyber’s SDK isn’t just a single innovation—it’s a stack of optimizations that work together to eliminate lag:
Real-Time Data Synchronization
- Combines video, audio, sensor data, and control inputs into a single stream.
- Uses adaptive bitrate algorithms (similar to video streaming) to adjust to network conditions.
Edge Computing Optimization
- Unlike traditional cloud robotics, Kyber minimizes reliance on distant servers.
- Processes data closer to the device (at the “edge”) to reduce latency.
Scalable Observability
- Monitors millions of devices in real time.
- Detects failures before they happen, critical for AI-driven fleets.
From VLC to Kyber: How Video Streaming Expertise Translates to Robotics
The Unexpected Connection Between VLC and Robotics
At first glance, video players and robots seem unrelated. But Kempf’s career reveals a common thread: real-time data transmission.
| VLC Media Player | Kyber’s Robot Control SDK |
|---|---|
| Optimized for smooth video playback | Optimized for sub-100ms robot control |
| Handles network fluctuations | Adapts to unstable IoT connections |
| Open-source, widely adopted | Designed for scalability (millions of devices) |
Why Existing Solutions Fall Short
Most robotics companies today build their own control systems—but these are often limited to small fleets (e.g., 2,000-3,000 vehicles). Kyber is designed for massive scale, where millions of devices need to be managed simultaneously.
Current limitations:
- High latency in cloud-based control.
- Difficulty pushing OTA (Over-the-Air) updates to large fleets.
- Lack of real-time observability for AI-driven systems.
Kyber’s advantages:
- Sub-100ms latency for safe operation.
- Edge computing reduces dependency on cloud servers.
- Scalable observability for AI-managed fleets.
The Physical AI Revolution: Why Kyber Is a Game-Changer
What Is Physical AI?
Physical AI refers to AI systems that interact with the physical world—think self-driving cars, robotic arms, or delivery drones. Unlike digital AI (e.g., chatbots), physical AI must process real-time sensor data and act within milliseconds.
Kyber’s role? It’s the “nervous system” for physical AI, ensuring that control signals, sensor data, and video feeds are synchronized with zero lag.
Use Cases: Where Kyber Could Make an Impact
| Industry | Potential Applications |
|---|---|
| Logistics & Delivery | Autonomous delivery drones, warehouse robots |
| Healthcare | Remote surgery, teleoperated medical devices |
| Automotive | Self-driving cars, remote vehicle control |
| Agriculture | Autonomous tractors, drone crop monitoring |
| Manufacturing | Industrial robots, quality control systems |
Why Lightspeed Invested $5 Million
Lightspeed Venture Partners, known for backing Anthropic and Mistral AI, sees Kyber as a critical enabler for physical AI.
“Physical AI is only as good as the underlying systems running it.” — Lightspeed Venture Partners
Kyber’s scalability and real-time performance make it a foundational layer for the next generation of robotics.
The Technical Deep Dive: How Kyber Works Under the Hood
The Core Components of Kyber’s SDK
Real-Time Data Pipeline
- Combines video, audio, and sensor data into a single stream.
- Uses WebRTC-like protocols for low-latency transmission.
Adaptive Bitrate Control
- Dynamically adjusts data transmission rates based on network conditions.
- Prevents buffering or lag spikes in unstable environments.
Edge Computing Integration
- Processes data locally (on-device or nearby edge servers).
- Reduces reliance on distant cloud servers, cutting latency.
Observability & AI Monitoring
- Tracks device health, network conditions, and performance metrics.
- Enables predictive maintenance for large fleets.
Benchmarking Kyber Against Alternatives
| Solution | Latency | Scalability | Edge Support | Observability |
|---|---|---|---|---|
| Kyber | <100ms | Millions of devices | ✅ Yes | ✅ AI-driven |
| Traditional Cloud Robotics | 200-500ms | Limited (thousands) | ❌ No | ❌ Basic |
| Custom In-House Solutions | Varies | Limited (thousands) | ✅ Yes | ❌ Manual |
The Future of Robotics: Why Kyber Could Be Everywhere
The Rise of “Robot-as-a-Service” (RaaS)
Companies like Amazon (with its delivery drones) and Tesla (with Optimus) are betting on mass robot adoption. Kyber could become the standard control layer for these fleets.
The Open-Source Question
VLC is open-source, but Kyber is currently proprietary. Will Kempf follow the same model?
- Pros of open-sourcing Kyber:
- Faster adoption by developers.
- Community-driven improvements.
- Cons of open-sourcing Kyber:
- Harder to monetize.
- Potential security risks if misused.
Potential Challenges
Security Risks
- Remote control systems are prime targets for hackers.
- Kyber must encrypt all data to prevent hijacking.
Regulatory Hurdles
- Governments may impose strict rules on remote-controlled robots.
- Compliance will be critical for global adoption.
Competition from Tech Giants
- Companies like NVIDIA (with Isaac Sim) and Google (with Robotics Cloud) are also working on robot control systems.
- Kyber must prove its superiority in real-world tests.
FAQ: Everything You Need to Know About Kyber
1. What exactly does Kyber do?
Kyber provides an SDK for real-time control of robots and drones, synchronizing video, audio, sensor data, and control inputs with minimal latency.
2. Why is latency such a big deal in robotics?
Even a 100ms delay can cause a self-driving car to crash or a drone to miss its target. Kyber aims for sub-100ms latency to ensure safe operation.
3. How is Kyber different from existing robot control systems?
Most solutions are built for small fleets (thousands of devices). Kyber is designed for millions of devices, with edge computing and AI-driven observability.
4. Who are Kyber’s competitors?
- NVIDIA (Isaac Sim)
- Google (Robotics Cloud)
- AWS IoT RoboRunner
- Custom in-house solutions (e.g., Tesla’s Optimus control system)
5. Will Kyber be open-source like VLC?
Currently, no. Kyber is a proprietary SDK, but Kempf hasn’t ruled out open-sourcing parts of it in the future.
6. What industries will benefit most from Kyber?
- Logistics & Delivery (autonomous drones, warehouse robots)
- Healthcare (remote surgery, teleoperated devices)
- Automotive (self-driving cars, remote vehicle control)
- Agriculture (autonomous tractors, drone monitoring)
7. How does Kyber handle security?
Kyber encrypts all data to prevent hacking. However, as a remote control system, it will need continuous security updates to stay ahead of threats.
8. What’s next for Kyber?
- Expanding partnerships with robotics companies.
- Improving AI-driven observability for large fleets.
- Potential open-source releases for parts of the SDK.
Conclusion: The Next Chapter in Robotics Has Begun
Jean-Baptiste Kempf’s journey from VLC Media Player to Kyber is a testament to how real-time data transmission is becoming the backbone of modern technology. While VLC made video streaming smooth, Kyber aims to do the same for robotics—at a massive scale.
With $5 million in funding and a clear technical advantage, Kyber is poised to become a critical infrastructure layer for the physical AI revolution. Whether it’s delivery drones, surgical robots, or self-driving cars, the future of robotics may well run on Kyber’s real-time control system.
One thing is certain: The robots are coming—and they’ll need Kyber to move smoothly.