How do I write my thesis with you?

First of all, thank you. If you are thinking about having my professor and me co-supervise your thesis, that means a lot to me. This is half a year of your work, and being on your shortlist is not something I take lightly. I will do my best to be worth the trust.

This is a collection of answers to commonly asked questions, meant as a quick overview with a few useful links. It is my informal read rather than any official position of my employer, so it may contain mistakes, and please tell me if you spot one.

1. Application

At ETH there is no central application or matching for a thesis or project. You find a professor willing to host you (or, in practice, a PhD like me, arranging it under one), and only then does any administration begin.

So the first step is less a formal application than simply an email to me.

(1.1) Please send me your resume and, more importantly, anything that shows what excites you, whether that is public work you are proud of (a paper, report, bachelor's thesis, or GitHub project), or a problem you would happily spend half a year getting to the bottom of. Nothing needs to be polished, and thank you for sharing whatever you have.

From there we would build a proposal together.

2. Proposal

The proposal is a pitch that argues our question is worth a paper at a top venue.

This is mostly my way of understanding whether I can be helpful to you. It is the part I care about most, and by the end we have quietly started something together.

(2.1) Before we write anything, we talk through the things around the question. What is drawing you to it, whether publishing is a goal for you, who we might want to collaborate with (I am fond of working with people from industry), and how we will source the compute your experiments need. These shape the project as much as the question itself.

(2.2) If you already have a topic in mind, wonderful, we start from there. If not, that is just as fine, and we build one together from what you sent me. Either way the work is the same, turning a rough interest into a concrete, scoped question that fits both your curiosity and what I can meaningfully supervise. It will likely change many times before it settles, and that is entirely normal.

When we write the proposal, two references guide us:

• ETH's pre-proposal guidelines
• How to write a scientific abstract in six sentences

Once my professor agrees, everything after that is just paperwork.

3a. Administration (for ETH students)

This section applies to you if you are already at ETH, where the thesis is registered inside your own department and the path is short.

(3a.1) Please register the thesis yourself in myStudies, under the research projects and theses section, giving the title, a start date, and both our names, with my professor as the main supervisor on record and me as your day-to-day advisor and co-supervisor.

Beyond that, two things to keep in mind.

• Duration and grading follow your department's rules, so please do keep an eye on the dates. In D-INFK, for instance, a Master's thesis is six months full-time and a late submission without good reason defaults to a fail. The exact forms and deadlines sit with your department's studies administration.
• If you are doing a joint degree, your home institution's rules take over.

3b. Administration (for external students)

This section applies to you if you are matriculated at another university and would like to do your thesis or research project with me at ETH.

Your home university owns the thesis, and ETH is simply where the work happens. The project is co-supervised from both sides. A professor there takes it on as part of your degree, examines it, and issues the grade, while the work itself is done in our lab, with me as your day-to-day advisor under my professor. Because your degree lives at home, the sequence starts on that side.

(3b.1) Your home university signs a recognition form confirming the project counts fully toward your degree and that you stay enrolled throughout.

(3b.2) My professor signs an ETH supervisor confirmation taking on the ETH side.

(3b.3) With both in hand, please apply through Mobility-Online with the usual documents, your ID, transcript, and proof of funds and insurance.

A few things are worth knowing before you begin.

• This is the project mobility route, for working on a single project, not an exchange, which is for taking courses and runs on a nomination from your home university.
• It is open to Bachelor's and Master's students from research universities, so not universities of applied sciences (Fachhochschulen).
• The stay runs between 91 and 365 days, on-site and full-time, and it is not paid.
• There is no tuition and no exchange scholarship, though labs can sometimes cover up to CHF 2,000 a month from their own funds where it genuinely enables the stay, so it is always worth asking.
• None of this earns you ETH credits, exams, or a transcript, and your degree and grade stay entirely with your home university.
• Exactly who signs on your home side, and whether a learning agreement is needed, follows your university's rules, so please do check that early.
• Please apply at least two months ahead, or three if you need a visa, since the visa tends to slip and retroactive applications are not accepted.

4. Working together

Once the paperwork is behind us, we have half a year to do research we are both proud of.

What follows is lightly adapted from Javier Rando's notes on supervising and running projects that really resonate with me.

• The work should be something that excites you. I usually ask students to suggest a bunch of topics they find interesting. I will try my best to make those happen or make suggestions that could turn vague ideas into research projects. I will also usually have a bunch of projects in my backlog that I am happy to share, but it is important that you work on something that excites you.

• You are the leader of this project. I will be guiding you through the process, but it is important that you feel comfortable and in charge of everything that is going to happen.

• You can do great research. This is not meant to be "Yahya has a research idea, and I am going to implement it and execute experiments". I want my students to do (or learn how to do) research. This is not easy, you will feel lost at times, but that is totally normal. It will feel like a rollercoaster, but we will get somewhere!

• The project will change 1000 times and that is okay. As we start working on a project, it is likely that we pivot several times or change the framing. Sometimes, we have even changed topics completely after several weeks. This does not mean the project is going wrong or that you are not doing a great job. This is extremely common and you should expect this to happen in great research projects.

• Let us align expectations early. We will try to get this research project as far as we can. We often publish projects in main conferences and/or workshops. It would be great if you share with me what is your ultimate goal (e.g. get a main conference paper for your PhD applications) so that we can try our best to make this happen.

• I am here to help you, so please ask anything. I am happy to help with anything: career questions, introducing you to other researchers, etc. Just ask! I will try my best.

• I make suggestions, you make decisions. You are the person working full-time in this project and know all the details much better than I do. I will make suggestions, which sometimes might be completely off. Please tell me when these are not good, and feel free to stick to only those suggestions that make sense.

• Some of the best ideas will be your own. You will come up with ideas. Sometimes they feel "crazy" or "stupid", but in my experience they often are great. Please share them with me and think through how to implement them. The process of turning ideas into experiments is extremely valuable!

• If you are blocked, please ask first. If you do not have compute, do not know what library/dataset to use, etc., please ask me first. We can often give you more compute, or I may know the solution and save you a lot of time. If I do not, I will tell you.

• Frustration is worth sharing. Sometimes experiments are not working out or you feel stuck somewhere, and that is okay. Please share with me what you think is going wrong, why things are not working, and we will work to fix those.

• Questions are welcome anytime. If you have questions, please send them as they come, rather than holding them until our meeting. I will get to those as soon as I can.

• Reminders are always welcome. I will sometimes forget stuff I promised I would give you. Feel free to send reminders frequently. You are not being annoying!

• Anything that can be handled async, ideally should be handled async. Simple/organisational questions are often better handled offline than in the meetings.

• I am generous with my time, so please be mindful of it. I usually give students at least 45min/week for meetings. Meetings should be productive. If you think there are no things to discuss (maybe you have been implementing code, or were sick that week) or those things can be handled async, feel free to cancel the meeting (I will not take this negatively!)

• Meetings go best when you come prepared. I think a productive meeting has the following structure. It is often a good idea to first present (1) the goal/agenda for our meeting, (2) what you need out of this meeting (e.g. today I want to run some results by you and then discuss which of these 3 experiments might be best to run next. Ideally, after this meeting I should know how to prioritize my next experiments and which infrastructure is best to run them). This will help us both steer the conversation and make sure we get everything done in time. Some students like to use slides, others use a GDoc, or others do not use anything. Feel free to organise the meetings however you feel more comfortable. If you want to present results, it is often good to select the interesting findings beforehand and compile them so that we do not need to navigate your code interpreter to find them.

• Qualitative results deserve attention. This is a general suggestion to do better research. Many insights come from observing qualitative results. We are often running experiments and only look at the numbers. Please take some time to look through the results and try to find patterns. For example, if you are measuring how good your model generations are, it helps to spend some time looking at good and bad generations, trying to understand what is going on beyond the metrics or how those fail. Also, a lot of our discussions during meetings will be about qualitative results!

• You do not need to prove productivity to me. In our meetings, please do not list the tasks you have done or justify why you did not work much last week. Every person manages their time differently and every week will be different. In our meetings, instead of focusing on tasks, I like to focus on research ideas (e.g. discuss results with me, questions, new experiment design, etc.).

• If you think things can be improved, please let me know! Each student is different and I try to adjust to their way of working. It is great if you can be transparent about things that you think can be improved in the process or that would make you feel more comfortable.

5. Defense

Once the thesis is written up, how it is examined depends on which side owns your degree.

(5a.1) If you are an ETH student, it is examined and graded here under your department's rules, and that grade is the one that counts, though whether there is a formal presentation or defense varies from one department to the next, so please do check yours.

(5b.1) If you are external, the examination happens on both sides at once, yet what ultimately counts is your home university's decision rather than ETH's. Some home universities accredit the defense you give at ETH and take the grade over, while others ask for their own, so this is a very individual thing, and the two sides run on different rules and rarely fall on the same day.

6. And beyond

The defense is not really the end, I hope. With some luck we have come close to what we set out to do, the paper we hoped for or another goal we set early on, and I gladly keep at it with you past the deadline.

Where it goes next is up to you, so for a referral into industry, a recommendation letter for a PhD, an introduction, or career advice, please just ask, and I give what leads I can.

My own master's thesis is one of my happiest memories, and I hope yours will be too.

So Long, and Thanks for All the Fish.