Member of Technical Staff - Firmware

Dimensional Job Post

Dimensional is building the foundation for physical intelligence in general-purpose robotics. We develop a unified framework that enables robots to perceive, reason, learn, and act across diverse real-world environments and embodiments. Dimensional's mission is to make physical intelligence programmable, scalable, and accessible -- accelerating the development of generalist robots that can adapt, learn, and perform reliably in the real world.

Our platform sits between robot hardware and AI models, providing a modular, composable architecture for training and deploying robotic intelligence  abstracting hardware differences and standardizing core capabilities—such as perception, spatial understanding, decision-making, and control—Dimensional allows developers and partners to build, transfer, and operate robotic skills efficiently across platforms.

You will own the layer between DimOS and the silicon. Today, the boundary between our software stack and the hardware underneath it — kernels, drivers, BSPs, vendor firmware — is spread across the team. This seat consolidates it: you make the hardware trustworthy, so that everything above it can assume timing, buses, and drivers simply behave.

This is a deep infrastructure seat with unusual surface area. One week you're writing a kernel driver for a new actuator; the next you're building a BSP for a new compute module, chasing a clock-domain bug in sensor timestamps, or reverse-engineering a sensor vendor's half-documented firmware protocol so the rest of the team never has to think about it.

• Design, write, and maintain kernel-level drivers for actuators, motor controllers, and sensors across our robot platforms.

• Build and own board support packages for the compute platforms we ship on (Jetson-class and beyond) — device trees, bootloaders, kernel configs, and power.

• Integrate and wrangle sensor vendor firmware — evaluating it, working around its bugs, negotiating fixes with vendors, and wrapping it in interfaces the rest of DimOS can trust.

• Own bus-level communication (CAN/CAN-FD, EtherCAT, UART/RS-485, SPI, I²C, Ethernet, USB) including timing, synchronization, and failure behavior under load.

• Maintain custom kernel builds: patching, configuring, and upgrading kernels with real-time requirements (-class) as the stack demands.

• Make firmware work CI-able: HIL rigs, replayed bus captures, and mocks so the distributed team can test hardware behavior without hardware in hand.

• Write production PRs into the open-source DimOS repo under the same review bar as every other engineer.

Must-haves

• 5+ years of embedded systems design and development in production environments.

• You have written kernel-level device drivers — for actuators, motor controllers, or comparable real-time peripherals — and shipped them. Not "modified an example driver": designed, debugged, and maintained your own.

• Real-time reasoning: latency budgets, scheduling and priorities, bounded queues, and a clear answer for what happens when a deadline is missed (degrade gracefully, never silently).

• Bus and driver depth across CAN, UART, SPI/I²C, Ethernet, USB — written or substantially debugged, not just used.

• Custom kernel experience: you've configured, patched, and built Linux kernels for embedded targets, and you can reason about what happens between interrupt and userspace.

• You have built your own BSPs — device trees, bootloaders, Yocto or Buildroot pipelines — for boards that shipped.

• Experience working with (and around) sensor vendor firmware: integration, protocol debugging, firmware updates, and vendor escalation.

• Instrument-level debugging: oscilloscope, logic analyzer, JTAG/SWD, kernel logs. You have diagnosis stories from below the API.

• Strong written English and comfort with fully async, distributed collaboration.

Nice to have

• Rust in production (our native layer is increasingly Rust), or embedded Rust curiosity backed by systems C/C++ depth.

• NVIDIA Jetson depth:
  Jet Pack/L4T, kernel customization, power modes, thermal tuning.

• PTP/time-sync across multiple compute units; time stamping discipline across clock domains.

• OTA/fleet update systems (Mender, RAUC, SWUpdate), secure boot, provisioning.

• Middleware internals — DDS, Zenoh, LCM, shared-memory transports.

• Board-level literacy: you read schematics, talk to EEs fluently, and can specify a connector unsupervised.

• Open-source contributions — kernel, U-Boot, Zephyr, Yocto layers, or anything we can read.