K

Kandou AI

Provides chip-to-chip AI connectivity solutions

Firmware Lead

Embedded EngineeringConfirmed live in the last 24 hours
Full-timeNo salary listed

About the Job

Requirements

  • At least 8 years of developing and debugging production real-time firmware in C/C++, including at least 2 years leading firmware modules, projects, or a small team.
  • Expert hands-on C/C++ experience in resource-constrained systems, with recent personal code delivery, code review, and complex defect resolution.
  • Firmware experience on a retimer, gearbox, PHY, AEC, AOC, optical module, networking ASIC, or comparable high-speed mixed hardware/software product.
  • RISC-V and/or ARM bare-metal or RTOS firmware experience, including interrupts, traps, CSR or register access, concurrency, and tight RAM/flash budgets.
  • RTOS development with Zephyr, FreeRTOS, or equivalent, including ownership of scheduling, synchronization, timing, and failure recovery.
  • At least one silicon or board bring-up cycle using JTAG and related debug mechanisms, with root-cause analysis across firmware, drivers, RTL, and hardware.
  • Experience with Python, Git/GitLab, Make/CMake, CI, unit and integration testing, and static analysis, using disciplined reproducible development workflows.
  • A BS in Electrical Engineering, Computer Engineering, or Computer Science, or equivalent experience.
  • Strong communication, planning, mentoring, and cross-functional leadership.

Responsibilities

  • Lead implementation and delivery of real-time embedded firmware for AEC retimer silicon across 224G/lane, 448G/lane, and forward-looking 896G/lane configurations while remaining a hands-on C/C++ contributor.
  • Implement and maintain boot, configuration, calibration, link-training, fault-recovery, and run-time supervision state machines on resource-constrained RISC-V or ARM platforms.
  • Translate approved firmware architecture into module designs, APIs, task and interrupt models, memory budgets, and hardware/software contracts, and surface implementation tradeoffs early.
  • Code, debug, and optimize firmware for SerDes equalization and adaptation, FEC-aware monitoring and re-adaptation, telemetry capture, thermal behavior, and power management.
  • Own low-level drivers and management interfaces including SPI, I2C/SMBus, I3C, MCTP, and JTAG, and integrate firmware services with the CMIS management layer.
  • Drive pre-silicon development and first-silicon bring-up, isolating issues across firmware, RTL, boards, and lab setups using JTAG, register traces, and targeted instrumentation.
  • Set coding and review standards; build unit, integration, and hardware-in-loop tests; enforce CI, static analysis, and release evidence; and improve Python-based build, debug, and test tools.
  • Break the roadmap into deliverable work packages, estimate and track execution, mentor firmware engineers, and coordinate with architecture, ASIC, DV, validation, SI, and applications teams.

Desired Qualifications

  • Experience with IEEE 802.3ck or 802.3dj link training and equalization, SerDes adaptation algorithms, channel-quality monitoring, and register-level PHY tuning.
  • FEC exposure including KP4 or RS-544, pre-FEC and post-FEC telemetry, symbol-error analysis, and FEC-aware recovery behavior.
  • CMIS 5.x firmware integration, module or datapath state machines, VDM telemetry, and sideband protocols including SMBus, I3C, and MCTP.
  • Experience with bootloaders, secure boot, root of trust, authenticated A/B firmware updates, rollback protection, and field-recovery mechanisms.
  • Pre-silicon firmware development using QEMU, SystemC, FPGA, simulation, or emulation; heterogeneous multicore or vector-processing systems.
  • PCIe or CXL experience, hyperscaler qualification, interoperability debugging, and production support for field or manufacturing issues.

About the Company

Kandou AI

Provides chip-to-chip AI connectivity solutions

Kandou AI designs and sells semiconductor connectivity hardware for AI workloads. It provides a memory fabric and CXL controllers that link processors to memory, enabling scalable, energy-efficient AI inference and training clusters built from affordable hardware. Using Copper MIMO interconnect, it reduces energy per bit and supports features like offloading AI training collectives, switchless memory for in-memory databases, and field-repairable co-packaged optics for better cooling. By targeting the data-movement bottleneck between processors and memory and offering chiplet-based fabrics and IP licensing, Kandou AI aims to make scalable AI infrastructure more accessible to hyperscalers, data centers, and system integrators.

Company size
201-500
Company stage
Late Stage VC
Total funding
$504.7M
Headquarters
Lausanne, Switzerland
Founded
2011

Recently Posted Jobs

Explore more opportunities at Kandou AI

See more jobs