Vellex Computing Snapshot
At Vellex, we empower critical infrastructure - from electric grids and industrial robotics to autonomous systems and wearable health monitors to make intelligent, real-time decisions directly from raw, physical-world data. By enabling computation at the source, we reduce latency, save energy, and unlock new levels of responsiveness and reliability. Our work is paving the way for a more efficient, resilient, and sustainable future powered by AI that is truly connected to the world around it.
Vellex is designing an analog compute platform that computes on raw sensor data, transforming the way machines perceive, process, and act on real-world information. Traditional computing forces the continuous nature of the physical world into discrete 1s and 0s, introducing energy overhead, latency, and memory bottlenecks. Vellex takes a fundamentally different path: it computes directly on the physical signals themselves, harnessing the laws of physics to perform mathematical operations with extraordinary efficiency. This approach unlocks a new class of ultra-low-power, low-latency computation ideally suited for battery-powered devices, edge intelligence, and mission-critical systems.
Founded by a team of Ph.D. researchers, engineers, and high-performance computing experts, Vellex was built on the belief that the future of computation lies beyond incremental improvements to digital processors. Digital scaling is reaching its physical limits—energy per operation, data movement costs, and thermal constraints are slowing progress in everything from smartphones to AI inference. Meanwhile, our world is becoming increasingly sensorized: cars, drones, satellites, factories, and wearables generate enormous volumes of continuous data that must be interpreted instantly.
Vellex emerged to solve this problem at the root. By operating directly on the currents and voltages produced by sensors, our platform avoids the costly analog-to-digital conversions and memory transfers that dominate energy consumption today. Instead of shuttling data back and forth between memory and logic, Vellex’s compute fabric performs operations in place, using the physics of the circuit to execute computation.
What We Do
Vellex builds programmable analog compute blocks paired with lightweight digital control. This hybrid architecture delivers the best of both worlds:
1. Analog for efficiency and real-time responsiveness
2, Digital for programmability, precision tuning, and system-level coordination
Our processors can handle tasks such as filtering, signal transformation, feature extraction, compression, anomaly detection, localization, and sensor fusion directly at the analog domain—before data ever hits the digital pipeline. The result is dramatic reductions in power consumption and latency.
Vellex offers a new way to build intelligent systems that are faster, quieter, lighter, more responsive, and far more energy efficient than what is possible with today’s digital-only solutions.
Who We Serve
Vellex collaborates with customers whose systems operate in environments where energy, size, latency, and reliability are critical. These include:
1. Autonomous and smart vehicles requiring microsecond-level reaction times for perception and control
2. Space systems and satellites where power budgets are severely constrained and radiation tolerance matters
3. Drones and robotics that demand high-performance onboard intelligence without sacrificing flight time
4, Robotics and Industrial predictive maintenance that rely on continuous sensing and real-time analysis
5. Smart sensors and IoT devices that must run complex algorithms on milliwatts
6. Defense and security systems where edge intelligence and low-signature operation are essential
7. Next-generation AI and data center workloads that require new methods to reduce energy-per-inference
Across these domains, our technology replaces expensive digital processing pipelines with physics-accelerated computation that dramatically improves performance per watt.
How We’re Different
Analog computation isn’t new—but Vellex approaches it with modern engineering, deterministic programmability, and a principled understanding of classical physics. We are not neuromorphic and not analog-in-memory computing. Instead, Vellex uses programmable analog circuits operating directly on sensor signals, enabling:
1. Predictable operation
2. Low noise
3. Stability across temperature/voltage/process variations
4. Compatibility with standard digital systems
5. Fully tunable compute graphs
This makes our platform practical and ready for real-world deployment—something previous analog approaches struggled to achieve.
Our strong IP portfolio, proprietary circuit architectures, and early lab-proven results create defensibility and a clear technology moat. More importantly, our approach does not require rewriting how industries think about hardware; instead, it slots naturally into existing chip and system development flows.
Founded by a team of Ph.D. researchers, engineers, and high-performance computing experts, Vellex was built on the belief that the future of computation lies beyond incremental improvements to digital processors. Digital scaling is reaching its physical limits—energy per operation, data movement costs, and thermal constraints are slowing progress in everything from smartphones to AI inference. Meanwhile, our world is becoming increasingly sensorized: cars, drones, satellites, factories, and wearables generate enormous volumes of continuous data that must be interpreted instantly.
Vellex emerged to solve this problem at the root. By operating directly on the currents and voltages produced by sensors, our platform avoids the costly analog-to-digital conversions and memory transfers that dominate energy consumption today. Instead of shuttling data back and forth between memory and logic, Vellex’s compute fabric performs operations in place, using the physics of the circuit to execute computation.
What We Do
Vellex builds programmable analog compute blocks paired with lightweight digital control. This hybrid architecture delivers the best of both worlds:
1. Analog for efficiency and real-time responsiveness
2, Digital for programmability, precision tuning, and system-level coordination
Our processors can handle tasks such as filtering, signal transformation, feature extraction, compression, anomaly detection, localization, and sensor fusion directly at the analog domain—before data ever hits the digital pipeline. The result is dramatic reductions in power consumption and latency.
Vellex offers a new way to build intelligent systems that are faster, quieter, lighter, more responsive, and far more energy efficient than what is possible with today’s digital-only solutions.
Who We Serve
Vellex collaborates with customers whose systems operate in environments where energy, size, latency, and reliability are critical. These include:
1. Autonomous and smart vehicles requiring microsecond-level reaction times for perception and control
2. Space systems and satellites where power budgets are severely constrained and radiation tolerance matters
3. Drones and robotics that demand high-performance onboard intelligence without sacrificing flight time
4, Robotics and Industrial predictive maintenance that rely on continuous sensing and real-time analysis
5. Smart sensors and IoT devices that must run complex algorithms on milliwatts
6. Defense and security systems where edge intelligence and low-signature operation are essential
7. Next-generation AI and data center workloads that require new methods to reduce energy-per-inference
Across these domains, our technology replaces expensive digital processing pipelines with physics-accelerated computation that dramatically improves performance per watt.
How We’re Different
Analog computation isn’t new—but Vellex approaches it with modern engineering, deterministic programmability, and a principled understanding of classical physics. We are not neuromorphic and not analog-in-memory computing. Instead, Vellex uses programmable analog circuits operating directly on sensor signals, enabling:
1. Predictable operation
2. Low noise
3. Stability across temperature/voltage/process variations
4. Compatibility with standard digital systems
5. Fully tunable compute graphs
This makes our platform practical and ready for real-world deployment—something previous analog approaches struggled to achieve.
Our strong IP portfolio, proprietary circuit architectures, and early lab-proven results create defensibility and a clear technology moat. More importantly, our approach does not require rewriting how industries think about hardware; instead, it slots naturally into existing chip and system development flows.