Skip to main content
FuriosaAI logo

Scale-Up Ethernet Engineer

FuriosaAI
November 17, 2026
Full-time
On-site
Hsinchu, Taiwan
Test Engineering Jobs, Level - Mid-Career

Job Title

Scale-Up Ethernet Engineer

Role Summary

Work on building and validating an Ethernet-based scale-up rack that connects many AI accelerators as a single system. The role spans endpoint and switch bring-up, cross-layer root-cause analysis, rack management, observability, and test automation to ensure stable performance under real AI workloads.

The team integrates hardware, firmware, switch NOS, management systems, and AI workloads to validate racks aligned with open standards (OCP/Open Rack) and to make rack bring-up and long-duration validation repeatable and reliable.

Experience Level

Mid-level (no specific years of experience stated)

Responsibilities

Deliver system-level validation and management for scale-up Ethernet racks across hardware, firmware, software, and fabric.

  • Validate scale-up Ethernet endpoints and SoC integration from pre-silicon prototypes through silicon; work with RTL and firmware engineers to resolve issues.
  • Configure and tune NICs and switches for target traffic profiles; validate link-layer behavior (buffers, queues, PFC, LLR, CBFC) to improve stability under congestion.
  • Perform cross-layer root-cause analysis using packet captures, device counters, logs, and telemetry to isolate faults to hardware, firmware, driver, software, or fabric.
  • Build clusters and run real AI workloads (collectives, LLM inference/training) to measure bandwidth, latency (including tail), and stability; detect regressions.
  • Validate link interoperability and stability across optical modules, DAC/AEC/AOC cables, and switch ports; use CMIS and module telemetry for verification.
  • Design and implement rack and fabric management that unifies node BMCs, switches, power shelves/BBUs, PDUs, and cooling with inventory, power sequencing, and firmware update capabilities.
  • Implement rack-level observability, failure detection, isolation, and RAS scenarios; verify recovery with fault injection and long-duration tests.
  • Assemble and validate full rack systems stage by stage, moving validation environments toward OCP-standard Open Rack configurations.
  • Automate remote console and power control, test execution, result collection, and dashboards for regression and stability validation.
  • Design foundations for AI agent–driven work orchestration (equipment reservation, remote-control APIs, job queues) and verify agent outputs against evidence.

Requirements

Must-have skills and abilities for day-one effectiveness, plus preferred skills that improve fit.

  • Must-have: Practical experience configuring and debugging data-center Ethernet (switches, lossless/RDMA) and network devices or systems on Linux.
  • Must-have: Experience developing or debugging semiconductor system software at register and memory-map level (SoC firmware, device drivers, silicon bring-up/diagnostic software).
  • Must-have: Hands-on bring-up, integration testing, or performance analysis of rack- or cluster-scale systems combining servers, switches, and accelerators.
  • Must-have: Ability to automate test execution, device control, and log collection with Python or shell scripting.
  • Must-have: Skilled at evidence-driven hypothesis construction and reproducible experiments using packets, logs, and performance counters for root-cause analysis.
  • Must-have: Regular use of AI coding tools/agents in day-to-day work and ability to verify their outputs against evidence.
  • Must-have: Strong collaboration skills and clear documentation of test methods and results.
  • Nice-to-have: Experience with 100G–800G link bring-up, MAC/PCS/FEC/SerDes, QSFP-DD/OSFP modules, DAC/AEC/AOC cabling, and CMIS-based module management.
  • Nice-to-have: Experience configuring lossless Ethernet and tuning congestion control (PFC, ECN, buffers, queues) and familiarity with switch NOS behavior (Broadcom Tomahawk experience helpful).
  • Nice-to-have: Experience with rack/server management (BMC/OpenBMC, Redfish, PLDM/MCTP, OpenRMC), OCP rack hardware (Open Rack), or building CI-based test labs and long-duration stability infrastructure.

Education Requirements

BS in Computer Science, Electrical Engineering, or a related field, or equivalent practical experience.


About the Company

Company: FuriosaAI

Headquarters: Seoul, South Korea

FuriosaAI develops high-performance, energy-efficient AI inference hardware and software. Founded in 2017 by semiconductor and AI engineers, the company builds AI-native compute platforms to reduce AI energy and operational costs and operates globally with offices in Korea, Silicon Valley, and an R&D lab in Lisbon.

FuriosaAI logo

Date Posted: 2026-11-16