
Embedded Linux Robotics Software Engineer
at Proception Inc — Making humanoids dexterous enough to thread a needle
What the role involves
Real-Time Systems, Cameras, and Robot Runtime Infrastructure
At Proception, we are building dexterous humanoid robotic hands for the next generation of manipulation models. Our hands combine high-DOF actuation, tactile sensing, visual feedback, and real-time control systems to help robots interact with the physical world with human-level dexterity.
This role is for someone who likes being close to the hardware, close to Linux, and close to the robot. You will work across embedded Linux, camera bring-up, real-time sensing, IPC, device drivers, and high-rate robotics runtime systems.
The Team
We are working on robotics, dexterous manipulation, embedded systems, and real-world AI deployment. We started Proception because we realized that humanoid hands are one of the biggest bottlenecks to building general-purpose robots.
The next generation of robot intelligence will not come from better models alone. It also needs better hands, better sensors, better data, and better real-time systems. That is what we are building.
What You Will Own
Instead of joining a large team to maintain legacy infrastructure, you will directly bring up and improve the systems that make our robotic hands work in the real world.
First 30 Days
You will
- Understand our embedded Linux stack, robot runtime, sensing architecture, and hardware bring-up flow
- Bring up or debug camera pipelines on Jetson Orin, RK3588, or custom embedded boards
- Work with V4L2, Media Controller, libcamera, device tree, and kernel logs to validate sensor paths
- Profile latency across camera, tactile, IPC, perception, and control pipelines
- Contribute fixes to real-time sensing and actuation services
First 60 Days
You will
- Improve reliability of high-rate data pipelines for cameras, tactile sensors, encoders, and control loops
- Optimize Linux scheduling, CPU isolation, memory locking, IO paths, and process coordination
- Integrate low-level drivers, kernel modules, or hardware interfaces into the robot runtime
- Build diagnostics for latency, dropped frames, IPC failures, sensor health, and system-level timing
- Help create hardware-in-the-loop test infrastructure for robot subsystems
First 90 Days
You will
- Own the embedded Linux camera stack for a prototype robotic hand or upper-body system
- Reduce latency and jitter in the tactile-visual-control feedback loop
- Support multi-camera synchronization, hardware triggering, and calibration workflows
- Build robust logging, monitoring, and field diagnostics for camera and sensing subsystems
- Help define the architecture for Proception’s real-time robotics runtime
Must-Haves
- Demonstrated ability to build reliable software on embedded Linux systems using C/C++ or Rust
- Strong Linux system programming fundamentals, including threading, IPC, scheduling, memory management, DMA, interrupts, and debugging
- Hands-on experience with embedded camera stacks such as V4L2, Media Controller, libcamera, MIPI CSI-2, device tree, or sensor drivers
- Experience bringing up hardware interfaces such as CSI-2, SPI, I2C, UART, CAN, EtherCAT, USB, or PCIe
- Ability to debug real hardware using tools such as gdb, perf, strace, dmesg, ftrace, logic analyzers, and oscilloscopes
- Experience building low-latency sensing, perception, actuation, or control pipelines
- Comfort working across software, firmware, electrical, mechanical, and robotics teams
Nice-to-Haves
- Experience with PREEMPT_RT, CPU isolation, real-time Linux tuning, or deterministic userland runtime design
- Experience with Jetson Orin, RK3588, Yocto, Buildroot, Debian-based embedded systems, or custom carrier boards
- Familiarity with ROS/ROS2, DDS, ZeroMQ, shared-memory IPC, GPU acceleration, camera calibration, stereo vision, or SLAM
- You Might Be a Fit If
- You enjoy debugging problems where the answer might be in the kernel log, an oscilloscope trace, a device tree overlay, a dropped frame counter, or a scheduler trace.
- You like robots, but you do not need a polished robotics framework to be productive. You are comfor