ML Algorithm Mapping and Performance Engineer, Core ML

About the role

Cerebras Systems builds the world's largest AI chip, designed for fast training and inference. The Core ML team develops algorithms for efficient large-scale training and inference, and is looking for an engineer who maps these algorithms to the Cerebras architecture, identifies when they outperform competing approaches, and studies how their advantages change as models, workloads and hardware scale. The role combines performance modeling, benchmarking and prototyping to characterize the efficiency frontiers of emerging ML algorithms.

Key responsibilities

• Build analytical and empirical performance models for state-of-the-art ML training and inference algorithms.
• Characterize asymptotic behavior and how trade-offs change with model size, sequence length, batch size, parallelism and hardware scale.
• Construct Pareto frontiers across model quality, latency, throughput, memory footprint, communication and compute cost.
• Develop prototype implementations and benchmarks for the Cerebras WSE and relevant GPU or software baselines.
• Analyze system behavior to find kernel, compiler, runtime, communication and algorithmic bottlenecks.
• Evaluate emerging techniques such as parallel token generation, diffusion and speculative decoding, attention, sparsity, mixture-of-experts, low-precision computation and distributed training.
• Partner with researchers and kernel, compiler, runtime, inference and architecture teams to recommend implementation directions.
• Develop tools and visualizations for performance projections and trade-offs, and share conclusions in reports, presentations and design reviews.

Required profile

• Bachelor's, Master's, PhD or equivalent practical experience in CS, computer engineering, EE, mathematics or a related field.
• Strong foundation in computer architecture, parallel computing and systems performance.
• Strong understanding of ML fundamentals and ML systems, including how model and algorithm choices affect compute, memory, communication, accuracy and scaling.
• Experience with performance modeling, complexity analysis, benchmarking or system simulation.
• Ability to move between mathematical analysis, experimental validation and engineering recommendations.

Required skills

• Proficiency in Python and comfort with C++.
• Experience profiling and debugging performance in ML, HPC, CPU, GPU or accelerator systems.
• Preferred: roofline analysis, CPU/GPU simulators, kernel optimization, hardware-software co-design, CUDA, Triton, PyTorch, JAX, transformer internals and research publications or patents.