ML Research Scientist - Inference ML

Cerebras Systems builds the world's largest AI chip, 56 times larger than GPUs. This architecture allows Cerebras to deliver industry-leading training and inference speeds; over 10 times faster than GPU-based hyperscale cloud inference services. This order of magnitude increase in speed is transforming the user experience of AI applications, unlocking real-time iteration and increasing intelligence via additional agentic computation. Cerebras works with the leading model labs, global enterprises, and cutting-edge AI-native startups. OpenAI recently announced a multi-year partnership https://openai.com/index/cerebras-partnership/ with Cerebras, to deploy 750 megawatts of scale, transforming key workloads with ultra high-speed inference. About The Role As an ML Research Scientist on the Inference ML team at Cerebras Systems, you will invent and evaluate advanced algorithms, methods and model architectures for efficient inference on our flagship Cerebras architecture. You'll work alongside ML researchers and engineers to design, prototype, validate, and optimize methods and models, contributing to cutting-edge inference research on the world's fastest AI accelerator. You will focus on pushing the frontier of speculative decoding, large-model pruning and compression, sparse attention, and sparsity-driven techniques to deliver low-latency, high-throughput inference at scale. Hybrid role in Toronto, ON, CA or Sunnyvale, CA, USA. Responsibilities - Research and prototype novel inference algorithms and model architectures that exploit the unique capabilities of Cerebras hardware, with emphasis on speculative decoding, pruning/compression, sparse attention, and sparsity. - Design and evaluate new transformer architectures and variants for NLP and computer vision to improve the trade-offs among quality, latency, throughput, and compute. - Identify and pursue high-impact research directions at the intersection of model architecture, inference systems and hardware. - Train or adapt models where needed, and measure and analyze inference performance using Cerebras tools to maximize throughput and minimize latency. - Collaborate with engineering and performance teams to understand practical constraints and help translate successful research into usable systems, from inception through delivery. - Publish and present research findings and contribute to open-source releases where appropriate. Minimum Qualifications - PhD in machine learning, computer science, or a closely related technical field with 3+ years of relevant research or industry experience, OR a master’s degree with an exceptional record of original, independent research and 4+ years of relevant experience. - Deep understanding of modern ML architectures and generative models in language and/or vision domains, machine learning systems, and the principles that govern their inference behavior. - Ability to reason about model and algorithm performance, particularly for inference workloads. - Ability to identify practically relevant research questions, develop novel methods, and evaluate their effects on model quality and inference efficiency. - Evidence of research impact in machine learning, statistics, information theory, or a closely related field, with clear relevance to modern ML, through publications in leading conferences or journals, high-quality preprints, or widely used open-source contributions. - Strong Python programming skills and proficiency with relevant ML frameworks such as PyTorch, Transformers, vLLM, or SGLang. Preferred Qualifications - Original research in one or more of: speculative decoding, neural network pruning and compression, sparse attention, quantization, mixture-of-experts models, sparsity, post-training techniques, or inference-aware design. - Track record of owning complex ML or inference projects end-to-end, from initial idea through iterative research, implementation, and evaluation to measurable impact in production. - Experience designing novel model architectures. - Strong foundation in performance optimization on specialized hardware (e.g., GPUs, TPUs, or HPC interconnects), with insights into how hardware characteristics can create new algorithmic opportunities. - Familiarity with large-scale model training and deployment, including performance and cost trade-offs in production systems. - Experience with Triton, CUDA, or C++ is a plus. Required Skills & Attributes - Self-directed mindset with a demonstrated ability to identify and tackle the most impactful problems. - Collaborative and generous with colleagues, with the ability to communicate ideas clearly across research and engineering teams. - Genuine passion for AI and a drive to push the limits of inference performance. Why Join Cerebras People who are serious about software make their own hardware. At Cerebras, we have built a breakthrough architecture that is unlocking new opportunities for the AI industry. With dozens of model releases and rapid growth, we’ve reached an inflection point in our business. Members of our team tell us there are five main reasons they joined Cerebras: 1. Build a breakthrough AI platform beyond the constraints of the GPU. 2. Publish and open source their cutting-edge AI research. 3. Work on one of the fastest AI supercomputers in the world. 4. Enjoy job stability with startup vitality. 5. Our simple, non-corporate work culture that respects individual beliefs. Find out more about what it's like to work at Cerebras here https://www.cerebras.ai/join-us! Apply today and become part of the forefront of groundbreaking advancements in AI! Cerebras Systems is committed to creating an equal and diverse environment and is proud to be an equal opportunity employer. We celebrate different backgrounds, perspectives, and skills. We believe inclusive teams build better products and companies. We try every day to build a work environment that empowers people to do their best work through continuous learning,