Compute Architect

MatX(11 days ago)

HybridFull TimeSenior$120,000 - $400,000Hardware Engineering

About this role

MatX designs hardware tailored for the world’s best AI models, building vertically integrated silicon-to-systems solutions to accelerate AI development. The company is hiring an ML accelerator compute architect to lead architecture work on next-generation AI chips. The role is based in Mountain View with a hybrid on-site schedule.

View Original Listing

Required Skills

Computer Architecture
Power Estimation
Area Estimation
Timing Estimation
Latency Analysis
Throughput Scaling
Numeric Formats
Quantization
Code Optimization
ISA Design

+6 more

About MatX

matx.com

MatX designs high-throughput chips and systems purpose-built for large transformer AI models, dedicating every transistor to maximizing performance for large-scale training and production inference. They prioritize performance-per-dollar and scale-out—supporting clusters of hundreds of thousands of chips and targeting workloads from ~7B up to 10T-class models, with competitive latency (e.g., <10ms/token for 70B-class). MatX provides low-level hardware control for expert users and aims to dramatically reduce cost and time-to-model so researchers and startups can train and serve very large models affordably and sooner.