'Tis the time of year for graduate applications

The same procedure as every year: The new academic year comes around, and universities open their application portals. Whether you are applying for a Master's degree or a PhD, or just frantically looking for funding, mentors and general advice, I'm going to walk through my personal experiences.

Every year, more people reach out to ask how I got here: how I applied, found my supervisor, fought through funding and scholarships, and stayed sane. My number 1 advice is still to reach out to as many people as possible and hear what worked (and especially what didn't!). But there's only so much time in a day, so I've written it all down here. And if you still have questions after reading, you can still reach out!

TL;DR

• Reach out to as many people as possible: supervisors, their students, people who've done it.
• Choose your supervisor for fit, not just the university's name.
• Course content beats rankings. Specialise early if you know what you want and experiment with your interests a bit.
• Time out to work isn't falling behind. It can make your application much stronger.
• Apply for funding early, and for more than one scholarship.
• Surround yourself with people who support you.
• When writing your personal statements: show your personal story, all your struggles and how you overcame them.
• Don't overdo it with GenAI/Agents. I know it's tempting to make the applications faster, but anyone who has to look at your application will spot it in a second because they get a bunch of the same applications. So be creative and authentic. Stand out because you wrote it yourself.

But let's start.

How did I get here?

I am currently a 3rd-year PhD student at the Language Technology Lab at the University of Cambridge, funded by the Bill and Melinda Gates Cambridge Scholarship. My work focuses on building community-centred language technologies (aka linguistically and culturally aware language models built by communities themselves and not big tech companies).

However, I started out in AI and Robotics. And before that, I started out as a Computer Science nerd.

If you are:

• Just curious? Start with Part 1: My story
• A current high school/secondary/college student thinking about your future? Go here
• Applying for an undergrad? Go here
• Applying for a Master's? Go here
• Applying for a PhD? Go here
• Just looking for funding? Go here
• Looking for general advice? Go here

Part 1: My story

First, how I got here. If you only want the practical advice, skip to Part 2.

School: How did I know CS was for me? And I was for CS?

Well, in year 8 at my school in Berlin, Germany, we had to select one course out of the following 4: Japanese (cool, but what are the chances I am actually going to use it?), Latin (I heard enough horror stories from my dad), drama (acting in a play was my own version of hell), and Computer Science (literally the only somewhat possible option). Little did I know that I would end up as the only girl in a room with 20 puberty-hitting gamer boys. I guess that's life 🙈.

But I persisted. And I started loving solving little problems and learning new programming languages. So I applied to the Hasso-Plattner-Institute (a university in Potsdam, Germany, which offered courses for high school students) to learn more about CS in my free time (don't worry, I had other hobbies, too 😅). I did this for two years, and it really widened my horizons, because my school curriculum could only cover so much. In between, I also started going to the Technical University Berlin, which offered free courses to students from the age of 16. Here I could choose which courses I wanted to sit in, do the homework and sit the exams, too. If I'd decided to actually do my degree there later on, I could have used these credits and shortened my studies. But alas, I left the country instead. But more on that later.

In school, I went out of my way to find extra resources; I signed up for any university or online course I could find. And over the years, there have been more and more resources to learn more about anything. I know I was privileged to be surrounded by two universities that offered free courses and a school that offered CS in the first place. If you are still in school and looking for extra resources, you'll find a few suggestions in Part 2 below.

Deciding to go for AI in my undergrad

AI was still a niche thing back in 2018; people weren't really talking about it outside of the nerdy CS groups obsessed with science fiction. I already annoyed everyone back then, and I knew that I had to study it. But usually you start with the basics: maths, physics, or CS in your undergraduate degree and then you specialise later on. With all my previous courses (at that point I'd gone to uni for 4 years already and I was just graduating from high school), I knew that I'd already covered most of the CS basics that I needed to specialise in AI. So instead of repeating all of that, I wanted to specialise in my undergrad already. I was very impatient.

Stubborn as I was, I looked for a university that already offered that. As I said, it was still a niche and not many universities thought that far. But I found the University of Bedfordshire, which offered AI and Robotics, which were all my favourite things in one. I was overjoyed. So I applied and got in.

Funding-wise, I received a bursary from the university, and I did an internship and continued as a Research Assistant at the German Aerospace Centre (DLR) to gain more work experience and pay off my student fees. Thankfully, my work connected to my studies, and I could write my final thesis in collaboration. So, if you're looking for work opportunities that also help you with gaining more experience in your field but also help pay off your fees, I'd really recommend looking at research labs in your area that offer research internships or research assistant positions. Sometimes you don't find it on their websites, so email the supervisors/PIs, or even some of the current PhD students to find out more.

Going for a Master's degree in AI and NLP

As you might have guessed, I'm a very curious person. So, after doing pure AI, CS, Robotics, Computer Vision, etc., I wanted to learn something new. Since my undergrad had already been so specialised, I couldn't really go for anything that wasn't AI, which was fine by me. So I applied to Queen Mary University of London, which offered a Master's in AI and an elective course in NLP. Of course, the AI parts were just repetition for me, but I went deep into NLP. Fortunately or unfortunately for me, ChatGPT came out right when I started my Master's, which meant that suddenly everyone was talking about AI and NLP. It really seemed like perfect timing.

Funding-wise, I applied to the Google DeepMind Scholarship, which runs at partner universities, including Queen Mary. It covered the fees, living costs, and offered a travel and technical equipment stipend.

The scholarship supports students from underrepresented groups in AI through Master's programmes at partner universities, Queen Mary among them. I first needed an offer for the course, then applied for the scholarship separately with a personal statement.

What I think made the difference: showing I'd already done research at DLR, being clear about why NLP and why now (and how it was personal to me), talking about being the only girl and person of colour in that year 8 classroom. Just sharing my journey, really.

Beyond the money, the scholarship came with mentoring from DeepMind researchers, events, and a cohort of other scholars, which really helped.

What to do next?

When I started my Master's (which was only a year long btw), PhD applications opened just two months in, and I had no idea what I wanted yet. ChatGPT had just thrown the whole field into chaos, and I didn't want to commit 3–4 years to a project that might be irrelevant by the time I started. I also wanted to learn more about NLP first, and to get some work experience outside academia. So I decided to wait and see. In that time, I noticed that people were so distracted by the wonders of chatbots that they ignored how these tools affected different groups in society. And with no ethics or social sciences in my studies, I realised I was missing valuable skills. How am I supposed to build technologies for everyone if I don't know how they interact with different communities? There's no one-size-fits-all.

I decided to go into policy for a bit. So I applied to a research assistant role at the Ada Lovelace Institute and got to work with other policy researchers for a year.

At this point, people usually ask: do you go straight into a PhD, or work first? (And what if you get used to the money?) For me it was a matter of timing, not doubt. At Ada I learned to run qualitative interviews, write for non-technical audiences, ask who specific technology is actually for, and by the end I'd found the rabbit hole I wanted to go down: community-centred language technologies. That year also made my PhD application much stronger, because my proposal was grounded in conversations with communities, policymakers, and civil society, and Gates Cambridge cares a lot about exactly that kind of impact. If you're worried that time out means falling behind: it doesn't. It gives you a clearer direction, and supervisors notice.

Applying for a PhD

So I'd finally reached the point I'd been waiting for so long: finding a place and some funding that will let me go wild and experiment on a topic I'm passionate about. It sounds great, but it's not easy to actually get what you want.

I had done all the textbook stuff: email supervisors, look for universities and funding, talk to current students, etc., but it still didn't get me a place. So I tried something else.

Someone had once told me that whatever you want to do in your PhD, it doesn't matter much if you are at the best university or have great funding if your supervisor doesn't treat you well or doesn't support you. I really took that to heart, since I knew I'd have a close relationship with a person in a position of power who'd hopefully guide me through at least 3–4 years of work.

Thus, I started by looking at potential supervisors who were working in my area first, and then also looked at whether our personalities would match (not that easy to track, but you can tell a lot about a person in the way they speak online).

Once I'd found a handful, I'd emailed them and also their students specifically. Of course, the return rate wasn't great, and I mainly got emails telling me to just apply and decide later. I ended up doing exactly that. I prepared a research proposal that was leaning towards the general goal/mission of the lab (in this case, equitable language technologies).

And then I got invited to an interview. In mid-January, I had a 13-minute interview with the two PIs of the lab. It went straight into my research proposal and questioned every single decision I'd made. So if you get to that stage, know your proposal inside out: not just what you want to do, but why you chose each part of it.

At the end of January, I got the offer. Cambridge was the only place I applied to. The supervisor and the research fit me so perfectly that it was either a PhD there or a completely different path.

Looking back, I think what made it work was the variety in my background. Having studied and worked on so many different topics made my portfolio very broad, so I could look at problems in an interdisciplinary way.

Part 2: The practical bit

Everything below is the how-to, grouped by where you are. Jump straight to your section.

If you're still in school

Looking for extra resources beyond your school curriculum? Here are a few suggestions:

• W3Schools: free, beginner-friendly tutorials for pretty much every programming language
• HPI-Schülerkolleg (Hasso-Plattner-Institut, Potsdam): a free programme for students in Berlin and Brandenburg from year 7 to year 12. It runs for a school year, with sessions every other week.
• Studieren ab 16 (TU Berlin): sit in on real university courses while still at school, take the exams, and get credits you can count towards a later degree there.

Free online AI courses you can take from anywhere:

• openHPI: the Hasso Plattner Institute's free online course platform, with courses on AI, machine learning, programming and more, mostly without prerequisites (some in German)
• Elements of AI: a free introduction to AI from the University of Helsinki, no maths or coding needed
• CS50's Introduction to AI with Python: Harvard's free course, once you can code a bit
• Google's Machine Learning Crash Course
• fast.ai: Practical Deep Learning for Coders, free and very hands-on
• Hugging Face courses: free courses on language models, NLP, agents and more

Not in Berlin? Many universities run similar schemes for school students, so search "[your city] university courses for school students" or "early study programme".

And look beyond courses, too: coding clubs, hackathons and girls-in-tech groups are great places to meet people who are as excited about this as you are.

If you're applying for an undergrad

Don't just pick the most famous university. Pick the course that actually teaches what you care about. I went to Bedfordshire, which isn't exactly a household name, because it was one of the few places offering AI and Robotics as a degree back then. That specialisation is what got me the internship, the thesis collaboration, and eventually the scholarship for my Master's.

A few things worth checking:

• Look at module lists, not just course titles. "AI" in a course name can mean anything.
• Check for placement years or industry links. They're the easiest route into paid, relevant work.
• Look at bursaries. Many universities have their own, and they're often less competitive than big national scholarships. Bedfordshire offered me one as a selected student, as a reduction on my fees.
• If you're international, check fee status early. I was lucky: as an EU student before Brexit, I paid home fees. That's no longer the case for most EU students in the UK, so check which fees apply to you before you fall in love with a course.
• Get involved in communities from day one. University societies, women-in-tech groups, hackathons, volunteering: they teach you as much as your modules, and the people you meet there will open doors for years (see Find your communities).

If you're applying for a Master's

• Think about what comes after before you start. UK Master's degrees are one year, and PhD applications open about two months in. If a PhD might be next, start thinking about it before you arrive.
• Use the electives. The core AI modules were mostly repetition for me, but the NLP elective changed my whole direction.
• Apply for funding at the same time as the course. Scholarship deadlines are often earlier than course deadlines.
• Go to conferences. Student volunteering and travel grants can get you in for free, and they're one of the best places to meet professors who are recruiting.
• Get involved in community activities. Join societies, volunteer, help organise events. It's where you'll find your people, and often your next opportunity (see Find your communities).

If you're applying for a PhD

Emailing supervisors (and their students!)

Most supervisors get dozens of these emails. Keep yours short, specific, and easy to reply to. Something like:

Also email their current PhD students. They'll tell you what the supervisor is really like: how often you meet, how much freedom you get, whether people are happy. That's what matters for the next 3–4 years.

If they reply, ask for a short call. It's the best way to check fit in both directions. Some supervisors already have a good idea of who they want to recruit before applications close, often people they've worked with or already spoken to.

Don't take silence personally. Many supervisors will just tell you to apply, and that's fine.

Writing the research proposal

Mine was around 1,500 words (the word limit) and roughly structured like this:

1. The problem: what's missing in current work and why it matters
2. Research questions: three main questions, one for each year of the PhD. I tried to be realistic about what fits into a year.
3. Approach: methods, data, and who I'd work with
4. Why this lab: how it connects to their existing work
5. Impact: who benefits

Nobody expects you to do exactly what's in your proposal. I framed mine as a broad umbrella project: AI changes so fast that I focused on the mission rather than specific details, which leaves room for the work to evolve. The proposal shows you can identify a problem, think it through, and fit into the lab. Lean towards the lab's broader mission rather than something so niche no one there can supervise it.

Want to see real examples? There's a community-collected archive of CS PhD statements of purpose. They're mostly for US programmes, but useful for seeing how people tell their story.

References and statements

• Ask your referees early, at least 4–6 weeks before the deadline. Send them your CV, your proposal, and a few bullet points on what you'd love them to mention.
• Mix is fine. I asked my undergrad supervisor, my supervisor at the German Aerospace Centre, and my manager at the Ada Lovelace Institute: two from research, one from policy. Non-academic referees work if they can speak to your research.
• Tell your path as a story. Mine went Bedfordshire → QMUL → policy → Cambridge. Each step should explain why you're ready for this PhD now.

Preparing for interviews

Interviews vary a lot, so prepare for a mix:

• Your past research. Be ready to explain what you did and why you made the choices you made.
• Your future interests. Have a clear idea of what you want to work on, even if it will change.
• Technical questions. You might get a maths or ML question, or occasionally a coding task (this seems to be more common in Europe).
• For Gates Cambridge, expect questions around the four criteria, especially leadership and how your work improves others' lives. Mine focused on the good my project could do: less on the technical side, more on people, ethics, and how my research could help communities.

A rough timeline

Dates are for 2027 entry; check the exact ones for your year.

Let's talk about money

Funding is often the hardest part. Here's everything I applied for and what I learned.

Google DeepMind Scholarship (Master's)

At Queen Mary, it's aimed at students from groups underrepresented in AI. For 2024–25 it covered tuition fees, a £17,000 living allowance, a £2,200 annual conference travel allowance and a one-off £1,700 equipment grant, plus a Google DeepMind mentor and invitations to AI conferences and DeepMind events. You need an offer for an eligible Master's first. If there are more eligible applicants than places, decisions are based on your personal statement and academic merit. Partner universities, eligible groups and amounts change each year, so check the current page (see Sources below).

ELLIS PhD Programme

ELLIS (the European Laboratory for Learning and Intelligent Systems) runs a PhD programme on top of existing university PhDs. You're co-supervised by two ELLIS researchers based in different European countries, and you spend at least six months with your secondary advisor during the PhD. There are academic, industry and interdisciplinary tracks.

One thing that surprises people: ELLIS itself doesn't fund PhD students, apart from smaller travel grants. Funding has to come from your advisors or your university. Mine is fully funded by Gates Cambridge. An ELLIS offer is also conditional on being accepted by the degree-granting university, so check whether your advisor needs a parallel application there.

You apply once a year through a central portal, choosing from a list of recruiting advisors. For 2027 starts, the deadline is 31 October 2026 and reference letters are due 25 November. To join, your main supervisor needs to be an ELLIS Fellow, and you can only be co-supervised by other ELLIS Fellows. My main supervisor at Cambridge is Anna Korhonen, and my co-supervisor is Desmond Elliott at the University of Copenhagen.

Gates Cambridge Scholarship

A full-cost scholarship for citizens of any country outside the UK doing a full-time postgraduate degree at Cambridge. Around 70–80 are awarded each year, roughly two-thirds to PhD students.

There's no separate application: you complete the Gates Cambridge funding section of your Cambridge application by your course's funding deadline. For 2027 entry that's 8 December 2026 or 6 January 2027, depending on the course (US citizens resident in the US had an earlier deadline of 14 October). Your department first nominates a small number of candidates, Gates shortlists from those for interview, and the final selection panels sit in late March.

The four selection criteria are outstanding intellectual ability, reasons for choice of course, a commitment to improving the lives of others, and leadership potential. The form asks four short questions, one per criterion. You also need a third reference, the Gates Cambridge reference, written specifically about those criteria. One of your academic referees can write it, as long as it's a separate letter focused on the scholarship rather than the course.

My experience: your Gates application depends on two things first, getting the PhD offer and being nominated (in my case, by my supervisor). After that come the four questions and the interview (see Preparing for interviews above).

Tip: the "improving the lives of others" part isn't a box to tick. Your whole application should show it, not just one paragraph. Community work you've done along the way is often the most convincing evidence.

Other options worth looking at

• University and departmental studentships, and doctoral training programmes

Apply for more than one. Rejections are normal. Funding is very competitive and often comes down to luck and timing.

Stay open to different avenues

My path doesn't look like a straight line. In my undergrad I worked on facial recognition, after seeing first-hand how badly it can work for people depending on their race or gender. Then I worked with a humanoid robot at the German Aerospace Centre, moved into NLP during my Master's, spent a year on immersive technologies and AI policy at the Ada Lovelace Institute, and now I work on language models for underserved language communities.

At the time, each switch felt a bit like starting over. Looking back, every one of them gave me something I still use: computer vision showed me how bias shows up in real systems, robotics taught me to build things that work outside the lab, and policy taught me to ask who a technology is for. My PhD topic only exists because I combined them.

So don't panic if you don't know your exact research area yet. Try things, take the internship in the field next door, say yes to the odd side project. You find what you care about by doing, not by planning, and the detours make you a better researcher.

Find your communities

None of this happened alone. Some of the most important support I've had came from communities that share my mission, whether that's getting more women and underrepresented people into AI or building technology that works for everyone. They gave me mentors, friends, opportunities I'd never have found on my own, and people to rant to when applications got rough.

If you're the only woman, or the only person who looks like you, in the room (I often still am), find these people. Join, show up, and when you can, give back. Here are the ones I'm part of:

Support networks for underrepresented people in AI

• Black in AI: a global community, free to join for AI practitioners and collaborators, working to broaden who gets to shape how AI is built, used and regulated. Its academic programme supports Black researchers applying to and navigating grad school, and it runs workshops and socials at major AI conferences.
• Black Women in AI: a professional association with members on five continents, built around educating, engaging, embracing and empowering Black women in AI through learning opportunities and mentorship.
• Women in AI: a global non-profit, founded in Paris in 2016, working to close the gender gap in AI through education, mentoring, events and research.
• Women in AI & Robotics: a German-Canadian non-profit network for women in AI and robotics, offering mentorship, workshops and hackathons, with local chapters.
• We and AI: a UK volunteer-run non-profit with a social justice mission, building critical AI literacy so more people, especially those most affected, have a say in if and how AI is used. Among other projects, it runs Better Images of AI.

Research communities that share my mission

• Masakhane: a grassroots, open research community strengthening NLP for African languages, for Africans, by Africans. The name roughly means "we build together" in isiZulu, and it's one of the best examples of community-led language technology there is.
• Centre for Human-Inspired AI (CHIA): Cambridge's interdisciplinary centre for AI that benefits humanity, bringing together researchers from across the University. I'm on its Early Career Community Committee.
• Ethics, Science, and the Public project: formerly the Kavli Centre for Ethics, Science, and the Public at Cambridge, it creates conversations between scientists and the public about the ethical questions raised by new discoveries.
• Centre for Computing History: a computing museum and educational charity in Cambridge, running workshops and events that bring the story of the Information Age to life.

Get involved: communities where you can do hands-on work

• Colorintech: a London-based non-profit working to make tech more inclusive for people from ethnic minority backgrounds, with career programmes for students and graduates, events, and ways to volunteer and mentor.
• Girls Who Code: an international non-profit closing the gender gap in tech, best known for its free coding clubs for school-age girls. You can join one, or volunteer to start and run a club yourself.

These are just mine. Look for what exists at your university, in your country and in your field, and if it doesn't exist yet, start it.

If you remember one thing…

Reach out. To supervisors, to their students, to people like me who've been through it. Almost every opportunity I've had came from someone taking time to answer a question, so I try to pass that on.

And look after yourself. Applications are exhausting: you're selling yourself over and over, often alongside a full-time degree or job. Reuse material (your proposal, statement and scholarship answers overlap more than you think), and remember that a rejection isn't a verdict on you. Have a cup of tea; it works wonders ☕️.

My path wasn't the textbook one: a university most people haven't heard of, a year in policy instead of going straight into a PhD, a research proposal built around what felt personal to me. It still got me here. Yours doesn't have to look like anyone else's either.

If you still have questions after reading all this, reach out. And good luck. You've got this ❤️

Sources and further reading

Scholarship details and deadlines change every year, so always check the official pages.

• Gates Cambridge Scholarship: gatescambridge.org and its FAQ
• ELLIS PhD Programme: Call for Applications 2026 and the ELLIS FAQ
• Google DeepMind Scholarship at Queen Mary: QMUL scholarship page (2024–25)
• Google DeepMind AI Master's Scholarships at other partner universities: IIE programme page
• CS PhD Statements of Purpose archive
• Abir Harrasse, On My Experience Applying to International PhD Programs: another great read, especially if you're applying from outside Europe or North America
• Free courses: openHPI, Elements of AI, Hugging Face Learn

Credits & disclosure

Header image: "Firehose II" by Pauline Wee, part of the Better Images of AI project. Pauline Wee & DAIR / Firehose II / Licenced by CC-BY 4.0

Parts of this post were drafted with the help of an AI assistant and then edited by me.

Comments

Name

Website

Comment

Post comment