Performance Engineer

About the role

Astera Labs (NASDAQ: ALAB) provides rack-scale AI infrastructure through purpose-built connectivity solutions. By collaborating with hyperscalers and ecosystem partners, Astera Labs enables organizations to unlock the full potential of modern AI. Astera Labs’ Intelligent Connectivity Platform integrates CXL®, Ethernet, NVLink, PCIe®, and UALink™ semiconductor-based technologies with the company’s COSMOS software suite to unify diverse components into cohesive, flexible systems that deliver end-to-end scale-up, and scale-out connectivity. The company’s custom connectivity solutions business complements its standards-based portfolio, enabling customers to deploy tailored architectures to meet their unique infrastructure requirements. Discover more at www.asteralabs.com.

Senior Performance Engineer

Location: San Jose, CA (On-site)

Role Overview

Astera Labs is a hyper-growth connectivity company enabling the rack-scale AI infrastructure powering the world's most advanced GPU clusters. Our Scorpio scale-up fabric switches are purpose-built to unlock the performance of next-generation AI workloads, and we're looking for a Senior Performance Engineer to demonstrate the real-world value of our silicon where it matters most: on real inference and training workloads running on GPUs at scale.

In this role, you will define how the world measures scale-up fabric performance. You'll build the roofline models, benchmarks, and end-to-end workload studies that quantify our performance leadership, expose bottlenecks, and drive performance fine-tuning of real AI workloads on our fabric to inform product direction. Your data will directly shape architecture, firmware, and product decisions — and fuel the marketing narrative that positions Astera Labs at the center of AI connectivity.

Key Responsibilities

- Performance Characterization & Benchmarking

- Establish theoretical and measured roofline models for Astera Labs' scale-up fabric across key performance metrics, defining the reference for all comparative testing.

- Build and maintain baseline performance benchmarks using industry-standard tools such as NVBandwidth and NCCL across a range of GPU configurations and switch topologies.

- Quantify the impact of differentiated Astera Labs AI fabric features (e.g., Hypercast, In-Network Computing) against baselines using both synthetic benchmarks and real inference workloads.

- Real Workload Analysis & Fabric Scalability

- Run end-to-end inference model workloads on target hardware to capture real-world performance beyond synthetic benchmarks, supporting architecture decisions and customer-facing demonstrations.

- Evaluate fabric performance as inference cluster size scales from 16 to 32 GPUs and beyond, identifying bottlenecks and building performance scaling models for state-of-the-art AI workloads.

- Design and execute head-to-head performance comparisons against competing fabric switch solutions to produce data-driven differentiation evidence.