Working with me

Working with me as a Post-Doctoral Researcher

Update Oct 2025: One (1) post-doc position available at Nangyang Technological University (NTU, Singapore), starting June 2026.

  • Topic: Bayesian Deep Learning: New Architecture Design in LLM Era
  • Requirements:
    • PhD in computer science/statistics or relevant AI/ML domains.
      • Have graduated/submitted your PhD by the starting date.
    • Good publication records in top ML conferences and/or statistics journals.
    • Aligned research interest - please don't waste your time contacting me if this is not the case - sorry!
      • Examples of aligned interests: Bayesian deep learning, uncertainty quantification, approximate Bayesian inference, attention/RNN architecture design, meta-learning & in-context learning (not pure LLM prompting type), online Bayesian inference/continual learning, speed-up methods for Bayesian computation and/or attention/RNN architectures, etc.
  • The position is based at NTU. Short visits to Imperial will be arranged subject to funding availability.
  • How to apply: Apply by emailing me your CV, and a link to your representative paper. Selected candidates will be requested for a reference letter from their PhD supervisor and invited for an interview.

Working with me for a PhD

Update Oct 2025: Several PhD opportunities available at two (2) universities due to various grant/scholarship arrangements.

You can apply one or multiple positions at one or both universities. However, if passing the interview, you'll only be accepted by one of the positions at one university. 

Subject to funding availability: short-term research visits will be arranged between universities (Imperial students to NTU or NTU students to Imperial).

I won't be able to advise on "vibe LLM prompting/agent" style research - sorry! Please don't waste your time contacting me if that's the case.

Imperial College London (Imperial, UK) - Jan 2026 / Oct 2026 entry

  • Position filled A time-limited opportunity, recruiting ends by Dec 2025, starting Jan 2026: AI4Health CDT studentship on "AI for pathogen deep-sequence analytics" (with Oliver Ratmann). Scholarship supports UK Home fee + 4 year stipend.
    • Recruited by Dec 2025 (via CDT admin).
  • At most one (1) General PhD (Oct 2026 entry): Topic is flexible but should be within my group's research interests.
    • We do probabilistic ML research, and a paper from us requires developments in both theory and scalable algorithms. See recent papers for examples.   
    • Scholarship is determined by the department's PhD scholarship panel, where I have little control of the process, and I can only nominate at most 1 candidate due to internal quota constraints. Typical winners for overseas students in AI/ML tracks have first-author full conference paper publications already.
    • Apply via university and submit required materials by Dec 15, 2025. List me as your potential supervisor. Selected candidates will be interviewed by the end of Jan 2026.
  • One potential CDT studentship (Oct 2026 entry) to be confirmed.
    • Details TBD.

Nangyang Technological University (NTU, Singapore) - Aug 2026 entry

One (1) PhD studentship for each of the following topics (fully funded):

  • Bayesian Deep Learning: New Architecture Design in LLM Era;
  • MLSys for Bayesian Computation (with Hongxiang Fan);
  • Causal Deep Generative Models for High-Dimensional Sequences.

Application process:

  • Email me your CV, transcript and a research statement, also include English and GRE test results when applicable. Selected candidates will be invited to submit official applications via university, and get interviews after checks by admission admin (on transcripts etc). 
  • The entire application process (including interview) will be finished by the end of Jan 2026 or early Feb 2026.

======

Before contacting me, think about whether you should (not) do a PhD in machine learning, read about my research topics, and think about whether we are a good match.

If you want to do a PhD with me then, I expect you to have a solid skill set in both math and coding. This might be demonstrated by a 1st-class UK degree (or equivalent) in math, physics, computer science or related subjects. I also provide a self-assessment here if you want a bit more reassurance. Publication is a plus factor but not necessary, as long as you can demonstrate your potential for research innovations.

When you contact me, attach your CV, plus a short statement (1 page maximum) about the potential ML research problems you want to work on with me and your expectations about PhD research. Also see the PhD application guideline for submitting an application to Imperial Computing.

Working with me as an MRes student (Imperial)

Update Nov 2024: We are changing the admission process for 2025/26 entry and onwards, interested applicants should just apply via Imperial application portal when it opens. Project selection will be done **after** admission.

Imperial Computing also has an MRes in AI & ML degree program where I am one of the supervisors there. This is a research master program where the degree will be awarded based on finishing a research thesis. Depending on the research results and interview, MRes students can continue for a PhD in AI & ML, or join AI industry for ML engineering roles.

Normally I expect to get 1 MRes students per year on projects proposed by myself and/or other faculty members in collaboration (i.e., you might get co-supervisors). I treat MRes students in almost the same way as 1st-year PhD students (the only difference is that they need to finish in 1 year), and I expect them to participate in the group activities of my research group. 

Please see the degree program website for application info. Once you get admitted, in the first month of the programme you can discuss with me on potentially working with me for your MRes project. 

Areas for PhD/MRes research topics

We can be a good match if you want to research on the following topics:

  1. Bayesian computation, e.g. approximate inference, sampling;
  2. Uncertainty quantification & decision theory;
  3. Training generative models for vision & NLP;
  4. Sequential generative models & stochastic dynamical models;
  5. Analysing behaviours of (stochastic) neural networks;
  6. Hierarchical/compositional representations, disentangled representations, generating counterfactuals, etc;
  7. Robust ML, e.g. security & privacy issues, model diagnosis/repair;
  8. Adaptive learning, transfer learning & continual learning;
  9. Personalised few-shot learning with e.g. knowledge representations;
  10. Others, e.g. energy-based models, gradient estimation, Stein's method, etc.

Short-term projects/internships/participation

If you are an Imperial undergrad/master/PhD student, contact me for potential projects. 

We will also host "crash courses" on advanced ML topics presented by our team members at the CSML reading group organised by people in Computing & Statistics departments, so reach out if you would like to participate. 

For other people, usually I do not offer paid short-term projects or internships. Still you are welcome to contact me for lightweight discussions and mentorships on related research topics. 

Collaborations

See here for a list of topics that we can potentially collaborate.