Senior Accelerator Engineer, Cloud AI/ML server team
Summary
Senior hardware engineer at AWS owning GPU/accelerator component lifecycle at hyperscale—defining qualification and firmware standards, leading fleet-wide failure analysis across hundreds of thousands of systems, and driving vendor technical relationships.
Application deadline: Sep 1, 2026
Do you want to become the go-to GPU expert for the world's largest Artificial Intelligence accelerator fleet, and grow that expertise faster than anywhere else in the industry?
This is a role where you will deepen your GPU expertise at a pace and scale that doesn't exist outside of hyperscale cloud, working directly with the industry's leading GPU vendors, analyzing behaviors across hundreds of thousands of accelerators, and defining standards that govern production deployment. If you have started to develop GPU or accelerator depth and want a position where you own the full lifecycle at fleet scale, this is the role!
AWS Hardware Engineering is looking for a Senior Hardware Development Engineer to own the technical roadmap across all AI accelerator platforms in the organization. You will be the single-threaded owner of GPU lifecycle from defining hardware, firmware and diagnostics requirements, qualification through field operations and translating fleet-scale failure data into vendor action.
Key job responsibilities
GPU Component Lifecycle & Strategy
- Own the technical relationship with vendors across all platforms in the portfolio: roadmap alignment, escalations, partnerships.
- Own qualification of new GPU SKUs and baseboard assemblies during NPI bring-up -- define test plans, acceptance criteria, and production readiness gates
- Define and maintain GPU firmware qualification criteria across the org -- pass/fail gates, staged rollout policy, regression detection methodology
- Drive RMA strategy: build failure evidence packages, negotiate acceptance criteria with vendors, manage submission quotas and pipeline velocity
Fleet-Scale Failure Analysis
- Lead root-cause analysis on fleet-wide GPU failure modes (component errors, PCIE interface errors, thermal events, link degradation, manufacturing escapes) using telemetry, event log data, and vendor diagnostics
- Set GPU health standards: define the metrics, thresholds, and alerting that platform teams execute against
- Define GPU fleet health dashboards: identify relevant telemetry, failure rate trends, replacement pipeline status, firmware version distribution, qualification status
Vendor Engagement & Cross-Team Leadership
- Represent the organization in technical discussions with leading vendor’s engineering -- translate fleet-scale patterns into prioritized vendor action items
- Partner with server teams to ensure consistent GPU operational practices; provide expertise without owning their execution
- Present GPU fleet health, replacement pipeline status, and qualification progress to senior leadership (VP-level) regularly
- Mentor engineers on GPU failure analysis methodology
A day in the life
No two weeks look the same. You might be engaging with our GPU vendor's engineering team on future roadmap options and how upcoming architecture changes affect our technical strategy. You might be defining technical requirements to enable AWS to optimize how we deploy and manage GPUs at scale -- translating fleet failure patterns into firmware feature requests. You might be analyzing thermal and error behaviors across tens of thousands of systems to develop predictive models that catch failures before they impact customers. Or you might be building the data package that proves a manufacturing defect to a vendor and recovers millions in component value.
What's consistent: you are the GPU component owner. You see every failure mode, every firmware release, every new SKU qualification. You develop expertise at a rate that isn't possible when you only see one system at a time -- here you see hundreds of thousands, and you use that scale to become the person both AWS and our vendors turn to for answers.
Located in Cupertino, Seattle, or Denver, you work with global hardware teams, vendor engineering, and cross-AWS accelerator initiatives.
About the team
AWS Hardware Engineering designs and delivers next-generation cloud infrastructure -- the servers, accelerators, and storage platforms that power AWS. Our team builds and operates custom AI accelerator systems at global scale, spanning GPU platforms from manufacturing through multi-year fleet operations. We are directly responsible for the most expensive and supply-constrained components in the AWS fleet.
Do you want to become the go-to GPU expert for the world's largest Artificial Intelligence accelerator fleet, and grow that expertise faster than anywhere else in the industry?
This is a role where you will deepen your GPU expertise at a pace and scale that doesn't exist outside of hyperscale cloud, working directly with the industry's leading GPU vendors, analyzing behaviors across hundreds of thousands of accelerators, and defining standards that govern production deployment. If you have started to develop GPU or accelerator depth and want a position where you own the full lifecycle at fleet scale, this is the role!
AWS Hardware Engineering is looking for a Senior Hardware Development Engineer to own the technical roadmap across all AI accelerator platforms in the organization. You will be the single-threaded owner of GPU lifecycle from defining hardware, firmware and diagnostics requirements, qualification through field operations and translating fleet-scale failure data into vendor action.
Key job responsibilities
GPU Component Lifecycle & Strategy
- Own the technical relationship with vendors across all platforms in the portfolio: roadmap alignment, escalations, partnerships.
- Own qualification of new GPU SKUs and baseboard assemblies during NPI bring-up -- define test plans, acceptance criteria, and production readiness gates
- Define and maintain GPU firmware qualification criteria across the org -- pass/fail gates, staged rollout policy, regression detection methodology
- Drive RMA strategy: build failure evidence packages, negotiate acceptance criteria with vendors, manage submission quotas and pipeline velocity
Fleet-Scale Failure Analysis
- Lead root-cause analysis on fleet-wide GPU failure modes (component errors, PCIE interface errors, thermal events, link degradation, manufacturing escapes) using telemetry, event log data, and vendor diagnostics
- Set GPU health standards: define the metrics, thresholds, and alerting that platform teams execute against
- Define GPU fleet health dashboards: identify relevant telemetry, failure rate trends, replacement pipeline status, firmware version distribution, qualification status
Vendor Engagement & Cross-Team Leadership
- Represent the organization in technical discussions with leading vendor’s engineering -- translate fleet-scale patterns into prioritized vendor action items
- Partner with server teams to ensure consistent GPU operational practices; provide expertise without owning their execution
- Present GPU fleet health, replacement pipeline status, and qualification progress to senior leadership (VP-level) regularly
- Mentor engineers on GPU failure analysis methodology
A day in the life
No two weeks look the same. You might be engaging with our GPU vendor's engineering team on future roadmap options and how upcoming architecture changes affect our technical strategy. You might be defining technical requirements to enable AWS to optimize how we deploy and manage GPUs at scale -- translating fleet failure patterns into firmware feature requests. You might be analyzing thermal and error behaviors across tens of thousands of systems to develop predictive models that catch failures before they impact customers. Or you might be building the data package that proves a manufacturing defect to a vendor and recovers millions in component value.
What's consistent: you are the GPU component owner. You see every failure mode, every firmware release, every new SKU qualification. You develop expertise at a rate that isn't possible when you only see one system at a time -- here you see hundreds of thousands, and you use that scale to become the person both AWS and our vendors turn to for answers.
Located in Cupertino, Seattle, or Denver, you work with global hardware teams, vendor engineering, and cross-AWS accelerator initiatives.
About the team
AWS Hardware Engineering designs and delivers next-generation cloud infrastructure -- the servers, accelerators, and storage platforms that power AWS. Our team builds and operates custom AI accelerator systems at global scale, spanning GPU platforms from manufacturing through multi-year fleet operations. We are directly responsible for the most expensive and supply-constrained components in the AWS fleet.