Annapurna Labs, an AWS organization with development centers in the U.S. and Israel, builds custom silicon and software for AWS customers. Our team combines cloud-scale innovation with world-class expertise across silicon engineering, hardware design, verification, software, and operations to tackle technical challenges that have never been seen before.
Join our Silicon Validation team to own power validation of next-generation machine learning accelerators that power AWS's cloud computing infrastructure. You'll work in a fast-paced, startup-like environment alongside some of the brightest minds in the industry on advanced, internet-scale technology that directly impacts how customers use Machine Learning acceleration. We are changing the landscape of cloud infrastructure by accelerating the development of custom silicon by moving beyond traditional partnerships to dominate in AI training and inference.
Your work will span power validation across the complete vertical stack — silicon power domains, DVFS flows, power delivery networks, dynamic workload power profiling, and system-level power budgeting. You'll validate power management schemes, characterize peak power and di/dt transient behavior, and close the loop between pre-silicon power models and silicon reality. Your measurements directly determine operating envelopes, influence architectural trade-offs, and define when silicon is ready for production deployment at AWS data center scale.
Key job responsibilities
As a Power Validation Engineer on our Machine Learning Acceleration team, you'll own power-domain validation across the entire product development lifecycle — from early design validation through emulation, silicon bring-up, post-silicon characterization, and ongoing support of production systems deployed in AWS data centers. You'll collaborate deeply with power architecture, RTL design, design verification, firmware, and software teams to ensure our next-generation AI/ML accelerators meet power targets and efficiency goals. This role requires bridging multiple domains — from low-level power delivery and analog measurement to workload-driven power-performance trade-offs — to deliver exceptional results.
We are looking for candidates with strong programming skills, computer architecture fundamentals, understanding of how workload behavior drives dynamic power consumption, a solid understanding of power architecture, as well as experience with power measurement, PDN characterization, and power management firmware.
A day in the life
- Developing comprehensive power validation strategies and detailed test plans covering sequencing, DVFS, peak power, di/dt, and stress testing from silicon bring-up to product release
- Characterizing peak power consumption and transient droop under real ML training workloads across PVT corners
- Conducting hands-on power measurements and debug in the lab using DC power analyzers, current probes, high-bandwidth oscilloscopes, and VNA/TDR
- Validating power management algorithms, DVFS transitions, thermal throttling, and power capping mechanisms on silicon
- Correlating silicon power measurements against pre-silicon power models and driving model-to-silicon feedback with architecture teams
- Building automated power regression frameworks, dashboards, and anomaly detection to track power across steppings at scale
- Collaborating across power architecture, design, firmware, and software teams to triage power anomalies and drive root cause analysis to closure
- Supporting production systems in AWS data centers and addressing power-related field issues as they arise
