My Journey to PhD Admission in Artificial Intelligence
It's a wrap!
After 6 intense months of college applications and anxiously counting down the days, I am pleased to say I'm pursuing my PhD in Artificial Intelligence this Fall 2023. I thought I would put together this blog to share all the valuable lessons I learned and experiences I encountered along the way that made all this possible.
Background
A bit of a background about me, I completed my Bachelor's degree in Robotics and Automation at a top private school in India. The school ranks 63rd nationwide and top 10 (out of ~500 universities) in the state. I took up coursework on Machine Learning and AI for Robotics, which sparked my interest in this field. While my time there shaped my academic success, I often wondered why I hadn't considered the path of research at that point. I have come across fortunate individuals who discovered their passion for research during undergrad, and actively engaged with professors, collaborating on works for peer-reviewed journals or conference proceedings. It's evident how publishing early on can be a game-changer, and I've witnessed how it has positively shaped the profiles of others in the field.
After my undergrad, I went on to work as a Machine Learning Engineer (MLE) at a startup. Although my time there was relatively short, I gained valuable experience working with a popular deep-learning framework – TensorFlow. I contributed to a project focused on the quantization of neural networks during the stages of training, evaluation, and inference. Through this process, I learned how to construct neural network architectures, modify graph layers, and create kernels from the ground up, allowing me to gain a comprehensive understanding of the underlying source code of TensorFlow. Working on projects (even personal projects) or roles like these will develop a strong sense of foundation of deep learning optimization techniques and enhance your thinking capability as a researcher. It will also provide the confidence to not only be the end user of a pre-trained algorithm but design custom or specific neural network models to suit your research objectives.
Although some skip this process of working in the industry (which is perfectly fine!), I personally saw the value in collaborating and gaining hands-on experience that gave me an edge when continuing to pursue my graduate studies. This process taught me how to effectively collaborate with teammates, how to drive projects from conception to completion, and how to fearlessly deep dive to make and break things in the process. Overall, my short stint as an MLE was a very fulfilling journey and laid the foundation for the next step of my journey.
No matter where you are in your journey, don't worry about the missed opportunities. Think about what you can control in the coming days, and put your effort towards the next most important thing.
As a next step, I progressed into enrolling in a Master's program aiming to delve deeper into AI. During this time, I enjoyed taking the extra leap in building a portfolio website, LinkedIn profile and tailoring my resume to showcase my abilities. Basically, build a professional brand. Look at yourself two years from now and keep adding things that make sense for future goals. But that's not all! This time also shaped my communication skills through various presentations, effectively allowing me to convey my ideas. Collaborating with professionals from diverse backgrounds helped me build a strong network that would last a lifetime. I also had several opportunities to work on exciting hands-on projects as part of the coursework, which helped me build a research portfolio and instilled a passion for research deeply.
I believe that passion for research is the curiosity about research's potential to push the boundaries of a niche domain while still making a positive impact in the community. Additionally, you get opportunities to collaborate with fellow researchers across universities – this teaches you the process of preparing for research lab meetings, critical thinking, and scientific writing. It also opens the door to attending technical conferences where one can engage with experts from diverse domains across the globe. This, combined with the freedom to explore an area of interest, is what I call an ideal "Researcher's World" – one where curiosity, resilience, collaboration, and impact intertwine. My research experiences fall under the larger umbrella of Responsible AI, and I aspire to be an expert in privacy-preserving AI/ML systems. What intellectually stimulates me every day is the process of reading existing literature, formulating research questions, and designing experiments to test my hypotheses.
During the second year of my Master's program, there were three pathways for me: coursework, project-based, and thesis-based. Right from the start, I made the decision to pursue a thesis, as it ensured that the path to a PhD would always remain open to me. The thesis is the most challenging path to Master's completion and is very similar to PhD dissertation but on a much smaller scale. I consider myself fortunate to have found a thesis advisor whose interests aligned closely with mine. I had gained interest in working with computer vision based generative models before the Generative AI buzz erupted in the industry. After months of immersing myself in the relevant literature, multiple iterations of carefully designed experiments, and 1:1 meetings with my advisor, I successfully completed my thesis on "Phoenix – A Federated Generative Diffusion Model."
Here are some key lessons I learned during this process (1) Stay relevant by keeping track of pre-prints and peer-reviewed works within your domain of interest (2) Be proactive in identifying opportunities and addressing gaps in the dynamic field of AI (3) Prepare for rigorous experimentation to test your approach thoroughly. For all Master's students aiming to pursue a PhD, here's my two cents: enjoy the journey of learning, and remember to seize every opportunity that goes beyond the curriculum.
PhD Application Process
Once I set my sights on pursuing a PhD in AI, I realized there was a lot to be done – with no clue where and how to begin the admission process. Being the first in my family to venture into this unknown territory, there was a lot of self-learning through blogs and videos, each of them detailing their personal journeys of applying for a PhD. The core essence of what I have understood and followed is explained below,
- Program – Dive deeper into the details and Research the program you would like to pursue (CS, AI, or EE). Then narrow down your research interests. This could be something that you were already familiar with (or) you could opt to venture into a new and exciting area that you are passionate about exploring further.
- University – There are no guidelines on how you choose universities, but individuals typically follow a number of university-ranking websites (like csranking, drafty.cs, usnews) and apply to a mix of dream schools, target schools, and safety schools. However, I would advise not to stick to the ranking list and make it the sole determining factor during the decision-making process. The research facilities, interdisciplinary collaboration, and funding opportunities mean much more. Remember, your life as a PhD student is heavily dependent on the quality of research and guidance from your advisor, which is more important than the university itself.
- Potential advisor – This is probably the most important factor to consider. Read your potential advisor's publications, and see if anything particularly piques your interest. Establishing a strong and supportive relationship for the next 4+ years of your life is crucial – allocate ample time during this process to thoroughly decide for yourself. Personally, I chose two advisors per university who aligned with my research interests. I also made notes of key points (Works of theirs I was impressed with, Why I'm a good fit, Our common goals and interests, etc.) that I could mention through personal communication and my application essay.
- Statement of Purpose – This will be your best chance of demonstrating your writing skills and highlighting your research accomplishments. Please don't start with generic phrases like "Ever since I was born, I wanted to do research." Be realistic and proudly describe your background in research, including when and where it originated. You should also provide a detailed explanation of your most notable research works, community involvement, and volunteer activities. Additionally, add key points on why you would choose to work with a particular advisor and how it could strengthen your research career. This essay should clearly articulate your strengths on why you deserve a place in their university/research lab to pursue PhD. I have seen sample essays that also highlight personal challenges and how they were overcome, but that's entirely up to you to decide whether or not you would like to include them in the SoP. While it's important to add all such technical details, don't forget to infuse your personal touch that best represents you. It's not a report!
- Recommendation Letters – You will need powerful recommendation letters from at least three professors. However, you could even request letters from two professors and an individual who is well-known in the industry. Although the content of the letters is confidential, you can request letters from individuals you have a strong work relationship with. By doing this, there is a greater chance that these letters will convey positive thoughts about your character and ability to perform research independently.
My Checklist
Throughout my phase of the admission process, I maintained a spreadsheet to stay organized. This could track the blogs I read, any sudden thoughts, and a comprehensive list of all universities that I intended to apply to. Here are some of the fields of the sheet, but feel free to customize it according to your need, as there isn't one template that fits all.
* University
* Department/Program
* Links
* Personal Ranking
* Deadline
* Tests GRE/TOEFL
* Application Fee
* Potential Advisors
* Recommendation Letters
* Extra NotesMost of the admission deadlines ranged from early December till the mid of January. To make the process more manageable and less overwhelming, I made it a point to apply to one university a day. Once the applications were submitted, the waiting phase began. This phase started with the temptation to check gradcafe, discord groups which made it all the more difficult to concentrate on everyday activities. I would recommend not getting stuck in this cycle like me since the results will gradually roll out after February. Around this period, you may also start receiving emails from professors expressing their interest in further conversations (like a match interview), whether in person or virtually.
Reflections and Lessons Learned
- Start the process early – research, research, and research! Not just about your publications but also about potential universities, advisors, and the domain of work.
- Build your profile – Collect all the artifacts right from undergrad and consider building a portfolio website to highlight your accomplishments.
- Networking – Leverage the power of Linkedin to connect with other individuals who went through similar paths. They often provide insights from personal experiences.
Good things take time and great things take a bit longer.
I hope this was helpful for those who want to get started with the process of applying for PhD. While it may sound tough and challenging, the process is worth it. Don't get swayed by peers or those around you who did not take this path. By choosing the road less traveled, the process is all the more exciting and requires immense passion and effort! Best of luck and I'll see you on the other side
