Flexible AI Hardware Research Engineer

graphcore; United Kingdom; Hybrid; GBP 75,000 - 120,000; Full time

Benefits offered by this job

Flexible working

Private medical insurance

Health cash plan

Dental plan

Pension (matched up to 5%)

Life assurance

Income protection

Parental leave

Employee assistance programme

Snacks and barista bar

Job summary

Graphcore is seeking an AI research engineer to advance AI/ML methodologies and translate ideas into scalable software-hardware co-designs. You will work with researchers and engineers on efficient training, inference, and hardware-conscious algorithms across multiple AI domains.

The role emphasizes publishing impactfully at conferences while delivering practical implementations that improve performance on Graphcore hardware and software stacks.

Qualifications

• Master’s or PhD in a technical discipline (e.g., Maths, Statistics, CS, Physics).
• Experience with Python in modern deep learning frameworks (PyTorch or JAX).
• Capable of designing, executing and reporting ML experiments; lower‑level programming for hardware efficiency (C++/CUDA/Triton).

Responsibilities

• Generate AI/ML ideas, design experiments, implement them & evaluate results.
• Prepare, submit & present work to AI conferences/workshops.
• Provide technical insight to internal teams by designing experiments and reporting results.
• Collaborate with researchers, silicon and software engineers to define, build and test AI hardware.

Skills

Python

C++/CUDA

PyTorch

JAX

Deep learning

ML experiments

Hardware-aware optimization

Education

Master’s or PhD in a technical discipline

Tools

CUDA

XLA/ATen

Triton

PyTorch

JAX

Job description

Graphcore is seeking an AI research engineer to advance AI/ML methodologies and translate ideas into scalable software-hardware co-designs. You will work with researchers and engineers on efficient training, inference, and hardware-conscious algorithms across multiple AI domains.

The role emphasizes publishing impactfully at conferences while delivering practical implementations that improve performance on Graphcore hardware and software stacks.