Intern - Research

Closing date: Rolling deadline

Location: London, Cambridge or Bristol

Start date: Flexible

Duration: 10 weeks (flexible)

About us
Graphcore compute systems are accelerating the AI revolution. 

Powered by the groundbreaking Intelligence Processing Unit (IPU), Graphcore delivers leading-edge AI performance with unprecedented efficiency.  

At Graphcore, innovation is everywhere. Our silicon engineers are doing pioneering work, including being the first company in the world to bring a Wafer-on-Wafer AI processor to market. Our software and AI teams are building new tools and optimising popular AI models, ensuring that we lead the market for performance, efficiency and usability. And Graphcore’s in-house research team is developing new approaches to AI compute, keeping our hardware and software ahead of the curve. 

The team

Graphcore Research is at the forefront of the Artificial Intelligence revolution.  Our team participates in both fundamental and applied research, to characterise the computational requirements of Machine Intelligence and to show the IPU driving the next innovative AI models.  We publish at leading AI/ML conferences (NeurIPS, ICML, ICLR) as well as specialist workshops in areas such as ML Systems, AI for Science, or GNNs, and collaborate with other research teams and organisations across the world.  Our team is based across London, Cambridge, and Bristol, with projects and discussions that involve all our locations.

The internship 

Acting as a member of the research team, you’ll investigate a research question that relates to machine learning or accelerated computation.  Practically, this means designing experiments, implementing models and algorithms, exploring data and presenting results. You will be largely self-directed in your research activity, with freedom (given appropriate mentoring and guidance) to pursue the most promising approach.  As part of the CTO group in Graphcore, you will present your findings back to the team, with the aim to present work publicly to the wider research community.

We are pursuing the following projects, each with exciting opportunities for interns at Graphcore, but are very open to other ideas that may align with the team’s goals — please get in touch if this is you!

• Accelerated computation in JAX — How can we develop/advance distributed programming models for modern ML hardware? (See: PySCFIPU (SynS & ML 2023),  Tessellate)
• Efficient transformer training & inference — Can sparsely accessed context push the limits of what is possible with transformer LLMs? (See: Unit scaling, PopSparse, GroupBERT)
• Knowledge graphs and graph neural networks — How can we improve the understanding and utilisation of graph-structured information in various domains? (See: OGB-LSC, BESS)

About You 
We’re looking for technically minded, highly motivated and adaptable problem solvers who can thrive in our collaborative team. Specifically, for this programme you’ll also need to: 

• Be a PhD student working in the field of Machine Learning and Artificial Intelligence
• Be able to evidence your coding skills (Python, C++), for example through projects or previous work experience, and experience of algorithm implementation in machine learning frameworks (PyTorch, JAX)
• An excellent communicator with strong team-working and interpersonal skills
• Highly motivated, action-oriented and a self-starter with a drive to get things done
• Passionate about the field of AI, both theory and practice, with a keen interest to continue learning

We are excited to hear about your research. Please provide a short summary of a recent project with your application (and we’d love to hear about any other relevant work).

Benefits 

In addition to a competitive salary our centrally located offices provide a well-stocked kitchen with healthy food, drinks and snacks and our very own barista (Bristol only for now, but we have great coffee everywhere). We have an active social scene too: cycling, yoga, running, board games, table tennis, football, to name a few.  

Additional information: 

Internships can be based in our Bristol HQ, Cambridge or London.

Working hours per week: 40 hours Monday to Friday with a 1-hour lunch break 

Unfortunately, no additional relocation support can be provided.