TixelJobs
P
Positron.aivia muse

Architecture & Performance Modeling Engineer, AI Inference Silicon (Senior / Pri

REMOTE$220K - $350K/yrPosted 6h ago
ML EngineerSeniorFull-time#performance-simulation-engineer#computer-architecture#hardware-modeling#silicon-architecture#senior-ai-hardware-engineer#remote

About the Role

About Positron AI Positron AI is building next-generation AI inference accelerators designed from the ground up for low-latency, high-throughput large language model inference. Chips built for the training era keep running into two walls at inference time: memory bandwidth and chip-to-chip networking. We designed our architecture to break through both. Our first-generation Atlas systems are already in production on Oracle Cloud Infrastructure. Asimov, our first custom ASIC, takes that architecture to silicon. In simulations, Asimov delivers several times more tokens per dollar and tokens per watt than leading GPUs, at a fraction of the power. Paired in our Titan system, Asimov is built to serve models beyond 16 trillion parameters. In September 2026, we raised $875 million at a $5 billion valuation. NEA led the round, joined by Atreides Management, Valor Equity Partners, and others, to bring Asimov and Titan to market. Our team includes engineers from Groq, Lambda, NVIDIA, Meta, F5, and many other industry leaders. Role Overview As a Senior Architecture & Performance Modeling Engineer, you will help shape the architecture of our next-generation AI inference accelerators. You will work directly with chip architects to explore architectural options, quantify tradeoffs, identify bottlenecks, and turn modeling results into actionable design recommendations. This work starts before RTL exists. You will use analytical models and our in-house simulators to answer architectural questions quickly, then refine and validate those models as the design matures. Each engineer on the team takes primary modeling responsibility for a next-generation chip while contributing to a shared modeling framework. Key Responsibilities Architecture Studies and Recommendations Partner with architects to turn design questions into focused modeling studies with clear assumptions, alternatives, metrics, and recommendations. Perform what-if analysis across compute engines, memory capacity and bandwidth, dataflow, DMA, command processing, on-chip interconnect, and chip-to-chip communication. Compare architectural options, explain performance sensitivities, and recommend changes within the system, power, area, and implementation constraints established with the architecture team. Workload Analysis Evaluate representative AI inference workloads across prefill and decode, batch sizes, context lengths, and dense and mixture-of-experts models to identify what limits latency, throughput, and hardware utilization. Provide timely feedback throughout the design cycle by communicating findings, investigating unexpected results, and updating studies as workloads and architectural choices evolve. Model Development and Correlation Develop and extend analytical, transaction-level, and cycle-level models, choosing the level of detail needed to answer each question with useful turnaround time. Maintain reusable models in our native Rust simulator and correlate predictions against RTL simulation, emulation, or silicon measurements as they become available, with regression coverage and clear documentation of assumptions, limitations, and discrepancies. Required Qualifications 5+ years of experience in computer architecture, microarchitecture, or performance modeling for accelerators, CPUs, or GPUs. Demonstrated experience using models to influence architectural decisions, such as identifying a bottleneck, comparing design alternatives, or recommending a change backed by quantitative analysis. A strong understanding of pipelines, queues, parallel execution, memory hierarchies, data movement, and interconnects. Hands-on experience building analytical, transaction-level, or cycle-level performance models, including first-order estimates that help validate and explain simulation results. Comfort working with incomplete specifications, making assumptions explicit, and balancing modeling fidelity with speed of iteration. Strong Rust skills, or expert-level C/C++ skills and a willingn
Share