What’s new?

  • Updated my PartII, III and Mphil proejcts (2022-2023) proposals (last updated 08/07/2022).
  • I drafted my PartII, III and Mphil proejcts (2022-2023) proposal for CS undergraduates at University of Cambridge, please check the link or contact me for details (last updated 08/06/2022).
  • My collaborators (Dr. Eiko Yoneki and Dr. Ilia Shumailov) are also looking for UROP students (co-supervised with me), feel free to send them an email if you are interested! (last updated 28/04/2022)
  • Please check this link if you are interested in having a UROP project (Research summer internship) with us. This is only for cambridge students. (last updated 16/04/2022)

Quick information for students

  1. I offer paid undergraduate summer internships, but they will have to be undergraduates from either Imperial College London or the University of Cambridge (I didn’t make these restrictions, the university did!).
  2. I can consider remote undergraduate summer interns, but these will not be paid due to UK tax laws.
  3. I have limited funding availabilities for PhDs, please see my notes below.
  4. If you are a student at the University of Cambridge and are interested in doing a Part II, III, or Mphil Project with me, please throw me an email, and see an incomplete project suggestions below (email me for more proposals or propose your own project!).
  5. If you are a student at Imperial College London and are interested in doing an Undergraduate or Master’s Project with me, please throw me an email, and see an incomplete project suggestions below (email me for more proposals or propose your own project!).
  6. It is worth to mention that my research interests are broad and not 100% hardware focused, I do a lot of work in the ML algorithm space focusing on AutoML, Security and Unstrctured data.

For PhD candidates

General advice

I do have limited funding availabilities for PhD students for the 2022 entry. I suggest the student look at the following general advice.

  • Applications for PhD are competitive. Unless you have a 1st class degree or a high 2/1 degree + a 1st class project, it is unlikely that you will be accepted by our group. MEng or BEng + MSc is preferable to BEng only.
  • Applications for PhD funding are more competitive.
  • If you have a BSc or a BEng from inside or outside Imperial College, you will need a good track record on publications.

You should always question yourself about ‘is doing a PhD worth it‘ before even going into an application process.

What I am looking for from you?

I am looking for students that are motivated and had demonstrated their studying skills in fields like Computer Science, Mathematics, Physics, and Engineering.

I expect students to have either a strong coding background or a strong math background.

Feel free to contact me if you are interested in joining my group.

PhD in AI/ML and What does it mean by working with me?

I will describe what does ML research mean to me and how I can help students in different ways.

  • I am actively writing code, in both software and hardware, and am happy to help people (including students) to debug.
  • I have a wide range of collaborations with researchers of different backgrounds and am always happy to refer students to my collaborators. I enjoyed doing a lot of placements/internships during my PhD and I would strongly suggest other PhDs to do the same given that they are on track.
  • I am not interested in helping you become an ‘engineer’ or ‘technician’, I would strongly encourage students to bounce ideas around and hopefully can become independent researchers at the end.
  • I am generally very chilled and am open to any research ideas.

My Research Interests

It’s important for you to understand my research interests, so you can know whether we are a good match.

I am interested in software, hardware, and security problems in the ML world, some more detailed past and current interests are:

Hardware Track

  • On-device ML: how can we make existing models run faster on our devices? Implementations of novel pruning, quantization methods.
  • Accelerated Machine Learning with custom computing: can we make ML models run faster using flexible computing units such as FPGAs.
  • Distributed training strategies and how to accelerate them on commodity or future hardware systems.

Algorithm Track

  • Automated Machine Learning: can we automatically find appropriate models given only the data and the task?
  • Data efficient ML methods: data-free and data-less optimization methods.
  • Learning with complex, unstructured data: information on graphs, hypergraphs and even more complex data structures.

Security Track

  • Finding new vulnerabilities in ML systems from the classic security engineering perspective.
  • Random cool topics on the ML security space.