Scale-Up Fabric Modeling Engineer for AI GPU Clusters

Astera Labs; San Jose (CA); On-site; USD 185,000 - 240,000; Full time

Job summary

Astera Labs is seeking a Fabric Modeling and Analysis Engineer to own performance models for the Scorpio Scale-Up Fabric. You will quantify bandwidth, latency, and throughput ceilings and predict behavior with AI/ML workloads before silicon exists, guiding architecture and roadmap decisions.

You’ll build models, calibrate against benchmarks, and collaborate with ASIC teams to drive end-to-end performance analysis and future interconnect innovations. This is a high-impact role at scale.

Qualifications

• Bachelor’s degree in Electrical Engineering, Computer Engineering, Computer Science, or a related field
• 8+ years of relevant background in network modeling, system architecture, and scaleup fabric benchmarking and analysis
• Strong programming skills in Python and/or C++ for building and analyzing system models
• Hands-on experience with cycle-approximate or event-driven system modeling and simulation
• Deep experience designing, enhancing, and maintaining fabric models using established frameworks such as Astera-Sim, Garnet, NS-3, ATLAHS, BookSim, or in-house simulators
• Solid understanding of high-speed interconnect and switching fundamentals, including PCIe Gen 6/7, ESUN, UEC or UALink
• Experience characterizing performance across bandwidth, latency, buffering, and congestion in network model or cluster

Responsibilities

• Simulation Infrastructure & Model Correlation
• Design, implement, and maintain features in AI/ML system simulators and fabric models, extending support for collective communication, transport, topology, congestion, routing, buffering, scheduling, and traffic management
• Improve simulator fidelity, scalability, debuggability, and runtime performance across packet-level, flow-level, and analytical backends, and build reusable abstractions and APIs for end-to-end simulation flows
• Continuously calibrate and validate models against benchmark data from the Performance Engineering team, owning the correlation between model predictions and measured silicon across scale up fabric generations
• Analytical & Fabric Performance Modeling
• Design and implement transaction-level or cycle-approximate system models of the Scale up fabric, enabling rigorous evaluation of new architectural ideas early in the design cycle, well before tape-out
• Develop and own theoretical roofline and analytical models that establish bandwidth, latency, and throughput ceilings for the Scale-Up fabric, mapping AI/ML workload demands against fabric capabilities to identify compute-bound vs. fabric-bound regimes
• Model AI/ML collective communication patterns (AllReduce, AllGather, ReduceScatter) and next-generation features such as advanced congestion control, In-Network Computing, and novel resiliency at scale
• Workload Characterization, Scalability & Topology Analysis
• Model fabric performance as clusters scale from tens to thousands of GPUs, evaluating topological choices and configurations and their impact on latency, bandwidth, and congestion
• Proactively identify architectural bottlenecks that would otherwise only surface at scale and propose solutions in the in-house network model
• Innovation, Competitive Analysis & Roadmap Influence
• Build analytical models of competing fabric architectures and evaluate emerging interconnect technologies — such as UALink, ESUN, and optical fabrics — to assess their impact on future Scorpio generations
• Translate modeling insights into architectural innovation proposals and patentable IP, and propose future switch features to architecture and engineering teams
• Serve as the internal authority on fabric modeling, partnering with ASIC architects and product teams to drive roadmap and feature decisions, support early customer engagement, and produce modeling reports, white papers, and presentations for engineering, product, and executive audiences

Skills

Python

C++

System modeling

Education

Bachelor’s degree in Electrical Engineering, Computer Engineering, Computer Science, or a related field

MS or PhD in Electrical Engineering, Computer Engineering, or Computer Science

Tools

Astera-Sim

Garnet

NS-3

ATLAHS

BookSim

Job description

Astera Labs is seeking a Fabric Modeling and Analysis Engineer to own performance models for the Scorpio Scale-Up Fabric. You will quantify bandwidth, latency, and throughput ceilings and predict behavior with AI/ML workloads before silicon exists, guiding architecture and roadmap decisions.

You’ll build models, calibrate against benchmarks, and collaborate with ASIC teams to drive end-to-end performance analysis and future interconnect innovations. This is a high-impact role at scale.