AI Research Engineer

About the role

Graphcore is seeking a technical expert to bridge the gap between machine learning research and hardware performance. You will develop new algorithms and evaluate them on our specialized silicon to shape the future of AI compute infrastructure. In this capacity, you will drive the discovery of novel model architectures and learning paradigms specifically tailored to data-parallel hardware. Your analysis will directly inform the microarchitectural decisions that define the capabilities of our processors. You will act as a crucial link between algorithmic innovation and the physical constraints of silicon execution. The position requires a mindset attuned to both theoretical rigor and practical engineering trade-offs. You will own the end-to-end validation of ideas from whiteboard concept to deployed hardware performance. This role is fundamental to ensuring that the software stack extracts maximum teraFLOPS from our intelligent processing units.

Key facts

What you'll do

• Conceive, execute, and analyze machine learning experiments to test new research concepts.
• Author and present findings at major scientific venues like NeurIPS, ICML, and ICLR.
• Deliver technical reports that influence internal product roadmaps and hardware design.
• Partner with silicon and software engineering groups to build and validate next-generation AI hardware.
• Profile end-to-end application performance to identify bottlenecks across the hardware and software layers.
• Implement custom kernels and operators to test the limits of programmable compute and memory hierarchies.
• Quantify the impact of architectural features such as vector and matrix engines on model training workflows.
• Explore the interactions between compiler optimizations, dataflow scheduling, and on-chip cache resources.
• Challenge existing assumptions by designing experiments that probe the boundaries of compute and memory bandwidth.
• Translate abstract research objectives into concrete metrics that guide the evolution of the hardware platform.

Requirements

• Master degree, PhD, or equivalent professional experience in a technical field such as Physics, Chemistry, Computer Science, Statistics, or Mathematics.
• Proficiency in Python using deep learning frameworks like JAX or PyTorch.
• Solid understanding of core deep learning concepts including scaling, model optimization, and evaluation metrics.
• Experience with low-level programming for hardware performance, specifically C++, Triton, or CUDA.
• Knowledge of deep learning hardware architecture, including vector and matrix engines, cache hierarchies, thread management, and data dependencies.
• Familiarity with software stacks, including kernel fusion, compilation, XLA/ATen operations, and asynchronous execution streams.
• Applicants must possess the legal right to work in the UK. Visa sponsorship is not available for this role.
• A demonstrated capacity to work effectively in a fast-paced, iterative research and development environment.
• The ability to communicate complex technical concepts clearly to both engineering and scientific audiences.
• Strict adherence to Graphcore's engineering standards for code quality, documentation, and reproducibility.

Nice to have

• Strong mathematical background in probability theory, linear algebra, and calculus.
• A track record of publishing research at international workshops or conferences.

Practical notes

• Applicants must possess the legal right to work in the UK. Visa sponsorship is not available for this role.
• Benefits include a matched pension up to 5 percent, private medical insurance, dental plan, income protection, life assurance, and a health cash plan.
• The company provides a flexible work policy, parental leave, and an employee assistance program.
• The Bristol office features an on-site barista bar and complimentary healthy snacks.

About the company

Graphcore has built a new type of processor for machine intelligence to accelerate machine learning and AI applications for a world of intelligent machines. The company was founded with the belief that current computing architectures are insufficient for the demands of modern AI workloads. By designing both hardware and software in tandem, Graphcore aims to unlock unprecedented levels of efficiency and insight for developers and researchers. The Bristol office serves as a critical hub for this innovation, hosting some of the brightest minds in computer architecture and machine learning. Success in this role requires a proactive approach to problem-solving and a commitment to excellence. The position offers the opportunity to work on foundational technology that influences the entire AI ecosystem. You will be expected to take ownership of your projects and drive them forward with minimal supervision. The fast iteration cycle means that your contributions will be visible and impactful within a short timeframe. This role is ideal for someone who thrives on challenging the status quo of AI compute. If you are passionate about building the hardware that will define the next generation of artificial intelligence, this position represents a unique chance to shape the future from the ground up. The work environment is collaborative, intellectually stimulating, and focused on delivering real-world engineering outcomes.