High Speed IO Validation Engineer

NVIDIA has been transforming computer graphics, PC gaming, and accelerated computing for more than 25 years. Today, we are tapping into the unlimited potential of AI to define the next era of computing — an era in which our GPUs act as the brains of computers, robots, and self-driving cars that understand the world. Doing what has never been done before takes vision, innovation, and the world’s best talent.

NVIDIA’s Silicon Co-Design Group is the team that gets every GPU, SoC, and CPU silicon program from first power-on to high-volume production. We are hiring an HSIO Validation Engineer to own characterization and qualification of high-speed I/Os — PCIe, NVLink, NVLink-C2C, UCIe, USB, UFS, and Ethernet — on brand-new silicon serving datacenter, AI, and client markets. Your bench is where silicon meets the signal — where every link train, jitter sweep, eye diagram, and protocol corner gets exercised before the part ships. You drive characterization and debug for IO interconnects across silicon, board, firmware, and driver — designing tests that anticipate rack-scale deployment so marginality surfaces at die-and-board scale, never in a rack. Your decisions show up in link margin, BER, power per lane, and the go/no-go calls that gate samples and production.

What you will be doing:

• Own end-to-end characterization and validation of high-speed IO — PCIe, NVLink, NVLink-C2C, UCIe, USB, UFS, Ethernet — from silicon die through package and board, on NVIDIA GPUs, CPUs, and SoCs.
• Develop and execute validation test plans for silicon and SoC-level HSIO features — bring-up, corner qualification, stress, productization, and structured DOE across PVT, board, cable/connector, thermal, and firmware variables — left-shifting silicon failures before deployment.
• Drive power characterization and optimization across IO interconnects under workload-representative stimuli (MLPerf-class patterns at silicon/board level) — analyze trade-offs across V/F, lane config, and protocol state — and feed back into design, firmware, and binning.
• Drive systematic root-cause analysis for complex HSIO bugs — isolating contributions across silicon, package, board, firmware, and driver, with platform/thermal/mechanical partners when system context matters. Convert rack-observed failures into silicon/board reproducers so future spins fix the root cause upstream.
• Assess manufacturing and yield implications of HSIO margin distributions — identify gaps in scaling from chip to product, and feed back into binning and design.
• Develop and improve validation methodologies, automation frameworks, and margin-tracking infrastructure — increase efficiency and coverage as platform complexity grows.

What we need to see:

• BS or MS in Electrical or Computer Engineering (or equivalent), with 8+ years in post-silicon HSIO validation, characterization, or a closely related discipline.
• Deep understanding of high-speed serial link protocols (PCIe Gen5+, NVLink, NVLink-C2C, UCIe, CXL) and the underlying SERDES architecture — equalization, link training, electrical characterization — with strong signal integrity and jitter knowledge.
• Strong SI and PI fundamentals — S-parameters, eye diagrams, channel margin, crosstalk, return loss, PDN noise, jitter decomposition — and validation experience across silicon and platform boundaries (CoWoS/interposer/substrate, board), translating rack-environment effects into silicon/board characterization.
• Hands-on with lab instrumentation (DSOs, BERTs, VNAs, protocol/logic analyzers, TDR), DOE/shmoo/margin-trending, and Python-based test automation — with a track record of driving cross-functional debug across chip, package, board, firmware, driver, and system teams.

Ways to stand out from the crowd:

• Hands-on experience with PCIe Gen5/Gen6 or NVLink validation on production silicon — link training, equalization tuning, BER margining, or compliance debug.
• Prior SERDES, mixed-signal, or analog circuit design experience — sufficient to independently form hypotheses, design Sim-to-Si correlation experiments, and propose silicon/system-side fixes during debug.
• Track record of translating rack-environment conditions (thermal, connector reliability, neighbor noise) into silicon/board characterization — finding marginality at die-and-board scale before it appears in the field.
• Comfortable using AI productivity tools (Claude, Copilot, or equivalent) to compress log triage, regression analysis, scripting, and reporting workflows.

#LI-Hybrid