PhD Applicants

A competitive PhD applicant is expected to demonstrate strong evidence of research potential.

Education: The prospective applicant should have a solid foundation in Computer Science, Electrical Engineering, Computer Engineering or closely related field as evidenced by previous degrees, coursework, or research experience. Research in the CHI-lab is at its core machine learning.

Research Experience: Applicants are expected to propose a compelling research topic or demonstrate an ability to contribute to existing research, showcasing critical thinking and a commitment to scholarly inquiry. The ability to contribute to existing research is evidenced through scholarly work in the form of at least one peer-reviewed scientific publication at a top-tier research venue, and as a first author. Going from a general idea to a research outcome is the most challenging when it is the first time you pursue it.

Professional Skills: Strong written and verbal communication skills, as well as the ability to work independently and collaboratively, are essential. Additionally, a PhD applicant should exhibit perseverance and a passion for contributing new knowledge to their field of study. These professional skills are evidenced through scholarly publications.

Emphasis on Research Productivity: While a strong academic background with high grades and prestigious awards can provide a strong foundation, it does not necessarily translate directly into the original thinking, meticulous attention to details, iterative effort, or the resilience required to navigate what is often a solitary journey, over an extended period of time. These skills are developed through practice and experience, rather than solely through academic achievements. The most reliable evidence of research experience is not the “time” spent in a lab, the “number” of lab rotations completed, or even the strength of recommendation letters. Instead, it lies in the validation of your scientific work through third-party peer review, such as the publication of research in reputable scientific journals. This process demonstrates the originality, rigor, and impact of your contributions, as recognized by experts in the field.

In the above context, the term “student” might be misleading; a more accurate description of the PhD program would be that of a junior research professional actively contributing to the scientific community.

Funding: PhD applicants that have or will secure external fellowships to fund their degree will be reviewed seperately from applications that have not secured external funding (i.e. seeking funding from the lab). Applicants that are seeking funding from the lab will be evaluated comparatively, unfortunately, there are typically more qualified applicants than there are lab-sponsored fellowships, so please consider this in your application process. Emirati applicants are encouraged to apply.

I hope this was helpful. Feel free to reach out to me with inquiries.

FAQ

What is a top-tier research venue in Computer Science / Machine Learning?

You can find a general list here. For a first-time researcher a venue in the top 100 may be considered top-tier. Typically each field has a set of venues that are considered top-tier, for example, for the NLP community it includes ACL, NAACL, EMNLP, for the Computer Vision community it includes CVPR, ECCV, ICCV, for general Machine Learning it includes NeurIPS, ICML. You may view many example publications from these venues through their proceedings (i.e. a digital documents / website that lists all publications accepted that year) or via scholar.google.com. Note that in Computer Science research is mostly disseminated via conferences (and not journals).

When should I start pursuing research?

I suggest starting as soon as you can. Many universities have programs where undergraduates can participate as researchers in a lab and be compensated for their time (e.g. 10 hours a week). This time is quite limited though, but at least you get exposure. Ideally, you will want to commit enough full-time attention (i.e. 40 hours a week) to gaining research experience towards a scientific research publication. The most successful PhD applicants I have seen will have started in their Sophomore (second) year and worked in a lab for 3 years (including Summers), and within a lab that has a record of training undergraduates to publish and pursue a PhD straight after graduation.

If you are seeking to do Machine Learning research, I also suggest you take an introductory course first, since it can save you from trying to learn on the job.

Publishing seems hard, shouldn’t I be learning this during the PhD program?

Yes, I acknowledge publishing the first time can be challenging, mainly due to a psychological barrier, that is, the process end-to-end has not yet been mapped by the ambitious researcher, and remains mysterious until they experience the whole process. In reality, a research paper is as short as 4 pages with a compelling table of results and a pretty figure. A research paper is a lot shorter than the dozens of reports a student will have composed during their undergraduate career. However, it does require a level of courage and critical thinking to walk through the mist to the other side. When I first started my own PhD program, I spoke to a faculty, which I had to find through a maze of oversized temporary MDF wall partitions, who described the research endevour as an endless fog, you wade through blindly, across “bones and skeletons”. I assure you, the reality is not that dreary, particularly if you can see past the mouse maze.

Should I pursue a Master’s (MS) degree first?

There are many paths, often non-linear, that one can take, and I’ve seen a number of successful individuals pursue such paths, including an MS degree. In general, CS-related MS programs tend to focus on training and upskilling of individuals interested in pursuing industry rather than a PhD, and will therefore have a large course load with limited research credits. Paradoxically, some PhD programs require students to have an MS degree before applying, therefore, I would encourage the prospective PhD applicant to carve out as much time as they can to improve their research profile, no matter what path they take.

One subtlety is that some structured learning through courses is helpful, for example, taking a Machine Learning or Signal Processing course ensures you are not learning fundamental concepts on the job, but other than fundamental courses, graduate courses are unlikely to be useful if you plan to conduct research in a rapidly evolving field (for which the knowledge base is fractured across research papers).