Introduction

A professor or the admissions committee cannot personally interview and speak to every student who applied; hence, the statement of purpose. Your statement should convey your motivation, enthusiasm, and, most importantly, the potential research directions you might want to explore during your PhD based on your previous experiences. Statements are expected to be about 1000-1200 words long, with a hard limit of two pages for the content. I understand that you might have a lot to share and mention in your SoP, but do not cheat by narrowing the margins or reducing the font size (I did that, and it does not look good). Just stick to Times New Roman with a font size of 11 and 1-inch margins on both sides. You can choose to write in a Google Doc, Word Doc, or use an Overleaf LaTeX template (to easily cite papers). The citations can flow beyond the two pages, and no one minds.

cs-sop.org is a good database of statements from previous students who received admits from various top universities. Give them a read, but do not overfit to them or unwillingly align yourself to any domain or problem mentioned there just for the sake of it. The whole admission cycle is about you, your learning, and your future ambitions, and it is okay (and expected) for different people to find different topics and research problems interesting. Similarly, there is no hard rule for the format of your SoP, except for the page formatting I mentioned earlier. You can choose to present your story in any way you wish, as long as it is technically sound, understandable, and genuine. Students usually cite their previous works in the SoP to convey their publication list and to connect the ideas presented and discussed with their prior work. What most professors reading an SoP look for is the relevance of your previous work, the connection to proposed ideas, technical soundness, alignment with their lab's research, and your enthusiasm and belief in your work. I understand all this is a lot to pack into 1000-1200 words, but that's what engineering is! Remember, you do not have to show off or force connections between all your works in the draft; your CV will take care of that.

More often than not, you may find yourself at a fork in the road where you want to highlight multiple disjoint works in your SoP. This can be a tricky situation, as the flow of the content may get disrupted. The best approach here is to discuss this with your mentors and try to explain your works under one cohesive umbrella, using all your learnings to align and corroborate them with your new proposals. Your SoP should not focus too heavily on your previous work and its technical details but should instead discuss your proposed ideas and hypotheses while clearly building upon your past research. That connection is the key!

In rare cases, when you want to switch domains, you need to be extra careful. Principal Investigators (PIs) usually prefer to hire students who already have some experience in their research area and can quickly adapt to the lab's pace. However, making such a transition is not impossible. There have been several cases where students have been recruited into new domains (including myself). You need to structure your SoP (and LoR) to convince the professor that you can learn their domain quickly and connect it to your existing fundamentals. It is best to familiarize yourself with the PI's recent publications and discuss in your statement how you can build on and improve their work. This helps give the PI confidence that they will not need to spend too much time teaching you the basics of their research. Your statement should also clearly highlight other technical skills and experiences that you can leverage to support such a transition.

I understand this might lead you to think, “Should I tailor my SoP to fit the PI's domain, or should I focus on describing my own work and ideas?” Often, professors—especially newer ones—are open to letting their students bring their own topics and work on them independently. However, established research groups with defined goals and funding may be more restrictive. Therefore, choose your PIs carefully based on your intentions. If you are strongly inclined to work with a specific PI, focus on their research and build your statement around it. But if you wish to explore and apply more broadly, go with the latter approach. The former can be slightly risky since your target space becomes smaller and other PIs may not connect with your proposal. That said, I have had friends who succeeded with both approaches.

Most SoPs end with a concluding paragraph where students are expected to mention why they are applying to a particular university and specific professors. Keep this section concise and to the point, and avoid spending too many words on it. Many students use a template where the technical content remains the same across SoPs for different universities and only the final paragraph is changed. Although this may save time, such forced stitching can break the flow for the reader. The biggest mistake one can make here is forgetting to update the university or professors' names in a rush (for example, applying to Harvard but leaving the last paragraph mentioning Stanford). Yes, this happens more often than you'd think! Although it may not look good, don't worry if this occurs—the admissions committee routes your application package to the relevant PIs. Every application process typically requires you to list 3-5 professors on the university's website when filling out your details. However, the best and most effective approach is to write separate SoPs for each university, focusing on their PIs and core research areas while uniquely connecting your work to theirs. It may feel like a lot of effort when you are applying to 15-16 universities, but no pain, no gain!

There are many excellent blogs by previous students who were admitted and have shared their SoP writing experiences in detail. I do not have anything drastically different to add, except that it is your story, and you get to decide how you wish to tell it. Any form of plagiarism is taken seriously and will lead to rejection. Revise and refine your SoP as many times as possible—it is a painfully iterative process. Try to keep the use of AI tools to a minimum and maintain a human touch. Get your SoP reviewed by as many people as possible, starting with your immediate mentors, who can help refine your problem statements and technical details. Next, share it with students from the university or lab you are applying to, so they can check whether your statement aligns with their group’s focus. In many cases, those students conduct the initial shortlisting of SoPs, so their feedback can be particularly valuable. However, remember that too many cooks spoil the broth. After a while, you may start receiving conflicting or overly specific comments. Be careful not to include anything that disrupts your original narrative.

UK and European universities often refer to the SoP as a research statement, which is typically longer and more detailed, spanning 3-5 pages. It includes an in-depth description of your problem statements and potential solutions. This document is almost like a short paper, where you aim to motivate the problem and present your hypothesis. Some professors later use it to apply for funding for your proposed project.

Make sure your SoP is modular to some extent so that you can condense or shorten it without losing much information or narrative flow. I remember UCLA had a text box with a 500-word limit for the SoP, unlike other universities that ask you to submit a PDF.

MS statements differ slightly from PhD statements, as they focus more on what interests you about the school. Since you are not targeting a specific professor for an MS, your statement should address the entire school or department and convey your motivations for pursuing the program there. This often blurs the line between a Statement of Purpose and a Personal Statement. Since I did not apply for an MS abroad, take this advice with a pinch of salt (or perhaps a bunch, as I like to say). My general understanding is that MS applications emphasize a Personal Statement, while PhD applications focus on a Statement of Purpose.

In addition to the Statement of Purpose, the university might also require you to submit a Personal Statement. This can be a bit ambiguous, as some universities use these terms interchangeably, while others refer to the SoP as a Research Statement or Research Proposal. A Personal Statement is typically a non-technical document of about 500 words describing you, your journey, life lessons, and motivation to join University XYZ for higher studies. This document should convince the admissions committee that you have not randomly chosen to pursue a PhD or MS at that specific University, but that you have a clear understanding of what you wish to learn, what you can contribute, how you align with the university's culture and values, and how this step fits into your long-term goals and ambitions.

If you are applying from India, universities usually require proof of English proficiency. The most widely accepted test scores are from TOEFL, IELTS, and GRE. However, this requirement can be waived if your previous university provides an official statement on your degree transcript or certificate indicating that English was the medium of instruction in that institution. You can upload this certificate and wait for it to be verified, which may take anywhere from a week to a couple of weeks. Choose carefully, as if this document is rejected by the university and no other English language proficiency score is submitted, your application could be desk-rejected.

Post-COVID, GRE scores for PhD applications are not considered but can be self-reported. So, if you do not need them, avoid taking the exam and save yourself the unnecessary expense and effort. Remember, all these tests are billed in US dollars and can be quite expensive in India. Additionally, scores must be sent from ETS (the organizing institute) to the respective universities, and ETS charges a fee for this as well. ETS frequently releases discount vouchers for exam registration and score reporting—keep an eye out for them, as they can save you a lot of money. However, for MS applications, GRE scores are still typically required and heavily weighted. A professor from Georgia Tech once told me that the first two factors considered in MS admissions are the GRE score and GPA. GRE scores remain valid for three years.

TOEFL is the most widely accepted English proficiency exam for PhD applications and is almost mandatory for all universities unless waived. The TOEFL consists of four sections—reading, writing, listening, and speaking—each graded out of 30. Every university sets its own per-section and overall cutoffs, with most expecting a speaking score of at least 28 for TA eligibility. Otherwise, you may be required to take an additional English course after admission. TOEFL scores are valid for two years.

IELTS is also an acceptable test; however, I have not seen many students take it (except that MIT considers it the primary metric), so I do not have much to say about it.

The most important advice I can give here is to start early, and I cannot emphasize this enough. If you are applying for an MS or PhD, aim to complete your exams by June or July to be absolutely safe, with August or September still being acceptable. October and November are the most crowded months, and you may not find available test slots. Moreover, you do not want to be distracted by test preparation during the crucial application phase. TOEFL scores usually take about a week to be fully processed, so plan a buffer for that. Additionally, linking scores to universities can take 3–5 business days, so account for that as well. It is highly recommended to take the internet-based test (iBT) at a testing center, as home tests are heavily proctored and prone to issues such as power or network disruptions, which could lead to your score being withheld.

One of the most crucial aspects of your application process is your connection with the professors you are applying to. It is highly encouraged to email them, let them know you are applying, and inquire about their recruitment status. In many cases, professors might be on leave, on sabbatical, under a hiring freeze, or about to move, so they may not be looking to take on new students. Since professors are often flooded with such emails during this period, your message might be missed or ignored—so do not be discouraged. You can send a polite follow-up if necessary, especially if you believe your research interests align well with theirs and their lab's focus, but avoid overdoing it. Keep an eye on their personal websites, where they often mention their hiring status and any specific instructions for prospective applicants. If a professor is not recruiting, do not waste time or effort pursuing them; instead, look for others who are actively hiring. At the end of the day, you want your application to reach someone who is open to taking students and can offer you admission.

I have also seen cases where mentors email professors on behalf of their students (or introduce them), which tends to have a much stronger impact since it serves as a direct endorsement. However, this is extremely risky for the mentor, as it puts their reputation and recommendation value at stake. As a result, many researchers openly refuse to do this. Only in very rare cases, when the mentor has a direct or personal connection with the professor and strongly believes the student is an ideal fit, might they consider sending such an email.

If you get the chance to attend a conference around the application period, go for it. Try to meet professors in person and share your intentions. Nothing beats a face-to-face, genuine conversation—you are far more likely to get a response and gain useful insights. Talk to their students to better understand the lab culture and the PI's expectations. I know of students who consistently stayed in touch with professors before, during, and after the application process. Many of them received good offers or were redirected to other professors who were actively hiring, which was equally beneficial.

This document answers the question, “How will the professor or admissions committee know that my work is genuine and that I did not fabricate my SoP or skills?” A Letter of Recommendation from your current or previous mentors provides the foundation for your SoP and formally introduces you to the reader. Many have told me that this letter is the most essential part of your application package and carries the greatest weight.

Every PhD application requires at least three Letters of Recommendation (LoRs). As a general rule, you should have one LoR from a college professor (Master's or Bachelor's) who can vouch for your academic performance. If you believe that LoR will not be very strong for any reason, you may choose to skip it. Most LoRs for an application are selected so that the reader is familiar with the recommender. This carries the most weight, as a professor is more likely to trust a letter written by a notable researcher, acquaintance, or close collaborator. Typically, LoRs are confidential—written and submitted privately by the advisor—but in some cases, the advisor might ask you to draft your own version, which they will then edit or submit as is. Discuss these details with your advisor beforehand—understand their expectations, how many LoRs they are willing to submit on your behalf, and plan accordingly. Share your SoP with your mentors so they can align their letters with your narrative. Your advisor may also ask you to specify what you would like them to mention. Try to ensure that each LoR highlights distinct aspects of your experience, showing different perspectives of your work and collaborations. This diversity is the main reason multiple LoRs are required.

An LoR can be strong or weak depending on how well the writer knows you and how confidently they can vouch for your abilities. The strength of an LoR is determined by the personal touch it carries and how closely the writer has worked with you. For example, if you took a course with a well-known professor and requested a letter solely for their name, the professor might write some positive but generic comments—which would still make the LoR relatively weak. Typically, recommenders do not write negative things about their students, but an overly generic, lukewarm, or unenthusiastic LoR is in itself considered weak and can hurt your application.

In the unfortunate case where you need to write your own LoR or a draft, here are a few tips. Remember, the LoR should describe your experiences but from the writer's perspective. You can enhance it slightly to highlight your strengths, but avoid exaggerating or portraying yourself as a “hero” student—it is easy to spot and will immediately reduce the letter's credibility.

• Your LoR should align with your SoP in terms of projects, learnings, and findings.
• Discuss your interactions—how inquisitive and driven you were, how you explored relevant literature and problems, and how you took ownership of the project.
• From the writer's perspective, describe your coding and implementation skills, how you overcame roadblocks, and how you continued contributing after project completion—such as driving open-source efforts or maintaining repositories and models.
• The letter should introduce you from the writer's perspective in the first paragraph, explaining who you are in the lab, how long you worked together, the topics you worked on, and the overall goals of the projects.
• The letter should convince the reader that you are capable of independent research during a PhD. My faculty advisor at IITM once told me that a strong LoR should sound enthusiastic—showing that the writer believes in your PhD ambitions and is happy to strongly recommend you to the university.
• Discuss projects in moderate detail, highlighting your contributions, publication venues, and any awards or notable achievements (for example, high citation counts or downloads). Mention any additional responsibilities you had with the writer, such as mentoring students, TA duties, or teaching assistance. These should appear at the end as supplementary information to highlight your overall skills. You may also conclude with your PhD aspirations, expressed from the writer's perspective if you previously discussed them together.
• Very important: Ensure that the LoR is dated within one month of submission and printed on official letterhead if possible, as some universities specifically require this. Send all LoR requests to the writer's official email address—some universities do not accept personal addresses or require additional verification. Try to keep the LoR 2-3 pages long; it should be detailed enough for the reader to understand why the mentor is confident in recommending you. Do not use AI tools to write it; keep the tone authentic and human. You may refer to sample LoRs and templates online, or use Overleaf templates if available, but ideally, the letter should use the mentor's institutional letterhead.

The application process can be quite expensive, with each university charging between 50 and 120 USD per application. Added to that is the cost of sending test scores to universities through ETS. However, many universities offer subsidies or fee waivers for students. In such cases, you may need to explain your financial situation and provide supporting documents to verify your claim. These fee waivers are typically limited in number and granted on a first-come, first-served basis. Verification of your case and documents by the university may take up to a week, or even longer as deadlines approach, so it is best to complete this process early to avoid a last-minute rush. Note that you do not need to submit your full application to apply for a waiver. The fee waiver process is handled independently, and you will be notified if you are approved, along with instructions for the next steps. Generally, you are not expected to submit the application until the fee waiver decision is finalized.

Additionally, ETS allows you to send your TOEFL or GRE scores to four universities for free immediately after the exam. So, memorize a few university and department codes in advance to link them on the spot and save money.

A CV serves as a quick introduction to you, your background, and your achievements. There are several CV templates available, and you can follow any standard format. Be sure to list your publications in a dedicated section. All achievements, accolades, and community service (not social work, but contributions to the research community such as co-organizing a conference, workshop, or tutorial, or reviewing for conferences) should go in another section. Your CV should also include a concise academic history, along with your GPA, any university rank achieved, and relevant work experience. You may choose to include one to three lines about notable, recent, or relevant projects, highlighting key findings and significant achievements. Avoid adding lengthy descriptions—this section is meant to give the reader a quick glimpse of your work.

A big mistake I made during my application cycle was not applying to safe schools. Safe schools are those universities where you are almost certain to get an admit, either because of your relatively strong profile or a prior understanding with a professor. Given the competition in AI and NLP, I strongly suggest maintaining a balanced gradient in your applications—around 20 percent top schools, 40–50 percent mid-range, and 30–40 percent safe schools. You can adjust this ratio based on discussions with your mentors or seniors who can help assess your profile. Applying only to top schools can be extremely risky due to their high standards and intense competition, which lowers your chances of admission. If you are absolutely sure you will be rejected by a particular university, do not waste your time, effort, and money applying there. Instead, focus on mid-range or safe schools. After all, your main goal should be to secure a shortlist or interview call.

Formally, any sort of preference or bias you introduce into your application process will narrow your sample space and, in turn, reduce your number of successful outcomes. MS and PhD applications are, to some extent, a game of probability. Suppose you have a success probability of p. If you apply to N schools, you can expect to receive offers from Np universities. Therefore, if you reduce N, your expected number of admits will also decrease.

The entire application cycle can be mentally exhausting, and you will see many people around you—who are also applying—feeling equally worried and anxious about interview calls and results. You will often hear discussions about who received interview invites from certain universities or professors you applied to, which can create a great deal of uncertainty. If possible, take a break during this period. Go home, relax, and disconnect for a while. Avoid engaging in anxious conversations if they are not relevant to you.