- Employment
- Full-time
About the role
Responsibilities
Design and develop the NPU Management Interface (MI) firmware/software enabling communication between Host/BMC and NPU devices
Implement and maintain MCTP, PLDM, and custom MI command handling for out-of-band NPU management, monitoring, and control
Develop device-management features over SMBus/I²C, I3C, PCIe VDM, or custom sideband channels
Integrate MI functionality into the NPU firmware, including:
Health and error reporting
Thermal and power telemetry
Runtime status, utilization metrics, and debug information
Ensure compliance with industry specifications by performing spec-driven design, implementation, and validation
Support bring-up, interoperability testing, rack-scale platform integration, and system-level validation
Develop test strategies and validation tools based on MCTP and PLDM specifications
Perform protocol compliance testing, regression testing, and interoperability verification
Requirements
Strong proficiency in embedded C or C++
Experience with firmware development for NPU/accelerator, GPU, or SoC
Understanding of management protocols including MCTP (over I²C/SMBus, I3C or PCIe VDM) and PLDM
Experience with low-level interfaces: SMBus/I²C, I3C, SPI, PCIe
Ability to interpret complex protocol specifications and convert them into robust implementations
Familiarity with device telemetry, sensor frameworks, watchdog/reset flows, and health monitoring
Experience with system-level debugging using logic/protocol analyzers and low-level debug tools
Knowledge of embedded systems, bare-metal or RTOS environments, and firmware lifecycle flows
Preferred Qualifications
Experience of BMC firmware stacks such as OpenBMC, Redfish, IPMI, and PLDM device-model implementations
Background in spec creation, requirement definition, or standards compliance validation
Experience defining FRU data, power/thermal management policies, and diagnostics frameworks
Familiarity with secure provisioning, firmware update mechanisms, and lifecycle state management
Experience with large-scale datacenter or HPC system integration (rack-level management, telemetry aggregation)
Contributions to firmware for accelerator, MCTP/PLDM implementations, or open-source system firmware projects
Contact
741,000+ hidden jobs like this
FuriosaAI and thousands of companies post here first — often days before LinkedIn or Indeed. Your first 5 applications are free; go Pro to apply without limits.
Everything Pro unlocks:
- Unlimited applications — free stops at 5
- Track every application in one place
- Apply straight to the source, one click
- Save & organize roles you love
- Roles pulled from company boards before the big sites
