Cerebras is hiring a ML Research Scientist Inference ML in Toronto!
Cerebras Systems, the company behind a wafer-scale AI chip substantially larger than conventional GPUs, is hiring an ML Research Scientist for its Inference ML team in Toronto. The role sits at the intersection of model architecture, inference systems, and custom hardware, with the goal of making large generative models faster and cheaper to serve on Cerebras accelerators. The researcher will explore techniques such as speculative decoding, pruning and compression, sparse attention, and other sparsity-driven approaches, then measure how well they hold up in terms of output quality, latency, throughput, and compute cost. Work spans NLP and computer vision transformers, and the position is explicitly research-oriented: publishing findings and contributing to open-source releases are encouraged. The team works closely with engineering and performance groups so that promising prototypes can become production systems. This is a hybrid position based in Toronto, Ontario.
Key responsibilities
- Invent and prototype inference algorithms and transformer architectures that take advantage of Cerebras hardware
- Investigate speculative decoding, model pruning and compression, sparse attention, and sparsity-based methods for low-latency, high-throughput serving
- Evaluate architectural variants for language and vision models across quality, latency, throughput, and compute trade-offs
- Train or adapt models as needed and profile inference performance using internal Cerebras tooling
- Identify high-impact research directions spanning model design, inference systems, and hardware characteristics
- Partner with engineering and performance teams to translate successful research into usable systems
- Publish and present results and contribute to open-source work where appropriate
Key requirements
- PhD in machine learning, computer science, or a related field with 3+ years of research or industry experience, or a master's degree plus 4+ years and an exceptional independent research record
- Strong grasp of modern ML architectures, generative models in language and/or vision, and ML systems, including how they behave at inference time
- Ability to reason quantitatively about model and algorithm performance for inference workloads
- Demonstrated research impact through top-conference or journal publications, quality preprints, or widely used open-source contributions
- Strong Python skills and hands-on experience with frameworks such as PyTorch, Transformers, vLLM, or SGLang
- Preferred: original work in speculative decoding, pruning, sparse attention, quantization, mixture-of-experts, sparsity, post-training methods, or inference-aware design
- Preferred: end-to-end ownership of complex ML projects and performance optimization on GPUs, TPUs, or HPC systems; Triton, CUDA, or C++ a plus
- Self-directed, collaborative, and clear communicator with genuine enthusiasm for pushing inference performance
- Seniority: Mid-senior (PhD + 3 years or master's + 4 years)
- Work mode: Hybrid (Toronto, ON)
- Key skills: Python, PyTorch, Transformers, vLLM, SGLang, speculative decoding, model pruning and compression, sparse attention, quantization, mixture-of-experts, transformer architecture, inference optimization, Triton, CUDA, C++
Want more jobs?
Join our mailing list to receive updates about new data science and ML Research Scientist Inference ML jobs.
