new students

It’s obvious that AI is changing the way hardware is designed but also creating an insatiable demand for new chips. I believe that it is time to raise the abstraction level, using only maths as the human-written specification and letting a combination of AI and new synthesis technologies determine the most efficient hardware. To give any trust in the hardware we will make formal proofs a first-class requirement of all of our tools.

My approach is built on three pillars:

  • Open-source EDA tools (e.g., CIRCT), as a platform for research exploration and industrial hardware design.
  • Equality saturation (and related methods) to massively expand the optimization search space.
  • Collaborative formal methods, combining multiple solvers to provide industrial-strength guarantees about the hardware we build.

Applying to work with me

I am actively seeking PhD students. Applicants should follow the official Imperial EEE PhD application process to understand when and how to apply.

Before submitting an application, please read some of my work to determine if we have common interests. Then I’d suggest reaching out with your CV and any research papers/projects you’ve already worked on. Please title any emails “Prospective PhD Candidate”. I appreciate personalised emails that demonstrate familiarity with my research. Generic or fully AI-generated emails make it difficult for me to assess mutual research interests.

What I look for

  • Intellectually curious - do you enjoy reading about a range of topics, do you ask questions in talks that are not in your field?
  • Independence - whilst I’ll be there to support, PhD students spend a lot of time working by themselves so you need to demonstrate that you’re self-driven.
  • Resilience - rejection and setbacks are part of the PhD experience, we should take on negative feedback and produce something even better.

Example PhD topics

  • Verified logic synthesis: bugs in synthesis engines can be fatal. We cannot blindly trust optimizations in open-source hardware tools, so how do we prove them correct? I’d like to explore hardware rewriting DSLs, solver technology improvements and proof orchestration of multiple solvers to address these concerns.
  • Flexible floating-point design: AI hardware increasingly mixes formats within one pipeline, but existing HDLs often express this as an explosion of muxes. Can we express multi-precision design more cleanly and generate optimized implementations automatically?

Expectations

What you can expect from me

  • A relatively hands-on supervision style, though unlikely to involve writing thousands of lines of code.
  • Weekly 30-minute meetings plus quarterly progress meetings with constructive feedback.
  • Responses within two working days, usually faster.
  • A publicly visible calendar showing when I am away.
  • Opportunities for outreach, networking, and regular conference travel.
  • Paper discussion sessions to build a clear narrative around where our work fits in the literature.
  • Interest in your broader goals, whether academia, industry, entrepreneurship, or something else.

What I expect from you

  • Clear and honest communication. If you’re facing delays or challenges, I would like to know as this lets me schedule my own time more effectively.
  • Responsiveness during working hours, and using the shared calendar to indicate time away from Imperial.
  • Dedication and effective time management. Every PhD student inevitably hits a wall in their research or takes on too many projects, so these need to be learning experiences.
  • Punctuality. Please don’t be late to meetings, especially if they are with outside collaborators.
  • Supporting the group. Students should expect to take on minor group administrative tasks, such as organising the group meeting, but also should expect to support other members of the group, particularly new students
  • To contribute to open-source research. Building tools keeps you humble and is the best way to deliver real-world impact. I want to be able to pick up any tool from the group and use it in a demonstration at any given moment.

What neither side should expect

  • Responses outside of normal working hours, if I choose to do so you should feel welcome to ignore my message until you return to work.
  • All the answers or results right away. Scheduling research is challenging and by definition we don’t yet have the answers only suggestions of how we might find them.
  • To keep up with all the latest literature. Research is moving faster than ever so it is impossible to keep up. We must be selective with what we read.

Group culture

My primary goal is to build a research group where everyone feels included and able to deliver their best work, no matter what their background is. In what follows I’ll set out the group culture I am hoping to build, and if this doesn’t align with how you work best then my group may not be the optimal fit.

Firstly, the default working location of all students should be in the office at Imperial and you should expect most meetings to be in-person. In my experience, this allows for high-bandwidth communication and collaborations to thrive. Of course, occasional home working should be deployed when needed.

My research is often inspired by industrial challenges, who we work closely with. You should expect opportunities to undertake internships during your PhD with chip design and EDA companies. I will create plenty of opportunities for you to build your own network, making introductions at conferences and taking you on industry visits, but what you do with those opportunities lies with you.

AI tools are of course of growing importance in research and I encourage students to treat them as tools, with a good level of scrutiny over what they produce so that students can explain and defend all aspects of the project. Presentations and communication are a big part of how I’ve got to my position so you should expect lots of opportunities to present internally and externally at conferences or companies.

I do not like leaving things till the last minute! I always strive to have deliverables, be that a paper or project report, ready several weeks before the deadline. Academic writing is a slow iterative process, where several rounds of feedback and refinement are required.

Throughout your PhD I will encourage you to retain “normal working hours”. That means getting to the office around 9am or 10am and leaving by 6pm, as this means everyone’s in the office together and people are maintaining a healthy work life balance. Research occasionally requires flexibility, but I don’t believe consistently working long hours is either necessary or desirable.

If we do have online meetings please have cameras on. Talking to a blank screen makes it much harder to build a working relationship and identify challenges you may be facing.

Our group is embedded in the wider Circuits and Systems research group in Imperial EEE, so we’re far from an isolated bunch and collaborations within the wider group are strongly encouraged. There is also strong overlap with the great researchers in the Computer Science department.

This is a living document and feedback is sought from all group members or potential students.

I really do believe that academics have the greatest impact through their students so, whilst we strive to deliver world-leading research, I’ll be most excited by what you go on to do in the future! If the vision laid out in this page excites you then I’d be delighted to hear from you.