My Approach to (Master’s) Thesis Supervision

Skills to Develop During a Master’s Thesis

There are many skills that contribute to a successful, and perhaps more importantly, fulfilling thesis experience. The list below is not exhaustive, but these are the themes I find myself reiterating most frequently in conversations with students. They can be grouped into three broad areas:

1. Foundations of Independent Research
   
2. Collaborative Thinking and Communication
   
3. Working Iteratively

---

Foundations of Independent Research

Asking a Good Question (and Letting It Evolve)

This is typically the first time you may not have predefined goals, unlike in most courses. Therefore, learning to think independently is a crucial outcome of a Master’s thesis.

A good research question is often S.M.A.R.T.—Specific, Measurable, Achievable, Relevant, and Time-bound. While not all research questions will satisfy this rigidly (especially in exploratory work), using this lens can help clarify your intentions and align your goals with what’s feasible within a thesis timeline.

Try to formulate an initial research question early and revisit it regularly as you learn more about the topic and discover the broader research landscape. It’s completely normal for the question to evolve as you dig deeper—in fact, refining the scope over time is a sign of progress, not failure.

Navigating Uncertainty Without Losing Momentum

Research inherently involves uncertainty. A Master’s thesis is a valuable opportunity to take something ill-defined and shape it into something actionable and concrete.

In fact, I would argue that learning to ask the right question is the ultimate learning objective—more important than finding a good answer. A clear, thoughtful question can unlock weeks of progress, while the wrong one can lead you in circles.

Always have alternate plans in place. Not everything will work out, and that’s okay. The real work is in learning how to frame a problem so it becomes solvable.

For many of you, this might be the first time you’re in full control of a complex, open-ended project. That can feel exciting, but also overwhelming. Unlike coursework, there’s no fixed syllabus, no step-by-step instructions, and no fixed outcome. This freedom is where real learning happens. But it also means we need tools and habits to manage the uncertainty.

Because research is full of ambiguity, I’ve found it helpful to borrow and adapt some principles from agile workflows from my time in industry, especially running teams at my former startup. While agile methodologies weren’t originally designed for academic research, their emphasis on iteration, prioritization, and feedback can be a powerful way to structure open-ended work like a thesis.

• Keep moving forward even when things feel fuzzy.
• Build habits of reflective planning and feedback.
• Learn from your peers (not just me).
• Turn “what do I do next?”” into a clear, answerable question.

None of this is about micromanaging you. It’s about creating just enough scaffolding to help you build your own rhythm, especially when the next step isn’t obvious. In the last section below, you’ll find a practical overview of how we put this into practice.

Practicing Epistemic Humility

The deeper you go in any field, the more you realize how much you don’t know, and how provisional many claims really are. One marker of a mature researcher is the ability to say “I don’t know” or “I am not sure if this would work” without anxiety. Expertise should lead not to overconfidence, but to curiosity and caution.

Be mindful of this when reading papers, evaluating your own ideas, or critiquing others. Ask: under what conditions would this fail? What alternative explanations might exist? Practicing epistemic humility makes you a better thinker, a better collaborator, and a more adaptive researcher.

---

Collaborative Thinking and Communication

Making Assumptions Explicit and Ideas Extendable

This is a skill that constantly evolves. It’s essential to keep iterating on how you communicate your ideas — in conversation, presentations, and writing. Clear communication not only helps you gain clarity yourself, but also enables others to give you more useful feedback.

One of the most important habits to develop is making your assumptions explicit. Even if something feels obvious to you, revealing the ‘why’ behind your choices—model selection, data preprocessing, experiment design—invites deeper, more constructive dialogue.

When brainstorming or discussing ideas with others, practice the principle of “Yes, And…”, which means accepting what someone offers and then building on it. I first encountered this in the Improvisational Acting course during my Masters at CMU’s Entertainment Technology Center—a program that brought together students from art, computer science, design, and engineering backgrounds.

In those interdisciplinary teams, “Yes, And…” wasn’t just a performance trick. It was essential for building trust and avoiding disciplinary snobbery. I’ve since found the same lesson holds in collaborative research: while we’re trained to be critical thinkers, scientists are often too quick to discard ideas at the first sign of a flaw. In my experience, however, early-stage ideation requires a different mindset, one that prioritizes momentum over perfection.

This mindset fosters creative exploration, mutual respect, and a more generative research process—especially when working across fields or with unfamiliar methods.

Giving Feedback That Builds, Not Breaks

While this comes with experience, practicing this during your Master’s can prepare you well for your future career. Start with noting what you like and don’t like about a paper when you read it.

Try asking: What assumptions does this method rely on? What baselines are included? What was unclear or surprising? Writing short critiques or comparative notes can help sharpen your judgment and give you vocabulary to discuss ideas with peers.

When giving feedback to others, keep in mind:

• Focus on the work, not the person
• Be specific, constructive, and actionable
• Offer ideas that move the work forward

The goal is not to “find flaws” but to collaborate on making the work stronger.

---

Working Iteratively

Learning to Share Before You’re Ready

Most work is iterative, and it takes courage to be vulnerable and share unfinished work to get early feedback. However, this is important for being efficient and productive.

Even sketching ideas visually or writing a ‘rough’ version of an idea can help clarify your thinking and allow others to help. Think of feedback as a process of externalizing and testing your assumptions, rather than a final judgment.

At the same time, try not to conflate the quality of your work with your value as a researcher—or worse, as a person. Everyone receives critical feedback. Everyone discards broken experiments. Research is hard, and progress is often nonlinear.

Instead, aim for excellence in your process, not perfection in your results. Let success be a natural consequence of good thinking, clear writing, and honest effort—while remembering that things like paper acceptance or reviewer scores are noisy signals. Learn to be proud of the work even when external validation doesn’t come immediately.

Prioritizing What Matters and Writing as You Go

Learning to scope your work realistically, and prioritize what’s necessary vs what’s nice to have, is critical in a time-bound thesis. This includes knowing when to stop tweaking an experiment and move on.

Documenting your reasoning and experiment setup as you go makes it easier to return to unfinished directions later and makes your final write-up less painful. You can find some useful resources on academic reading and writing here.

My Supervision Style and Philosophy

While I’ll adapt to your working style, I believe it helps to be transparent about my general approach, so we can begin with aligned expectations.

I believe that my role is to facilitate your growth. Consequently, I take a largely hands-off approach because I believe the deepest learning comes from active exploration. That means giving you room to experiment, even get stuck, because working through uncertainty builds the kind of intuition that instruction rarely provides.

That said, I won’t leave you adrift. I’ll nudge you when needed, offer perspective, and suggest resources—but I expect you to take the lead. To borrow a pottery metaphor: you put the clay on the wheel, and I help shape it.

This process works best when we see supervision as a partnership—one that balances autonomy with reflection, and challenge with encouragement.

Working Together: Tools, Platforms, and Weekly Rhythm

I suggest a lightweight, transparent system inspired by agile research practices. The goal is not to burden you with bureaucracy, but to give you just enough scaffolding to:

• Plan and prioritize your work
• Reflect on your progress
• Share ideas with your peers
• Identify blockers early so we can support you

To do this, we use three main platforms:

• Mattermost: for async check-ins and lab-wide discussion
  
• GitLab: for managing research tasks, ideas, and weekly planning
  
• GitHub: for version control, collaborative coding, and open science

This setup is designed to strike a balance between institutional constraints, collaborative transparency, and research freedom. GitHub remains the canonical home for your codebase (easier to make public if published), while GitLab provides a powerful yet simple interface for tracking ideas, issues, and progress across your research process.

---

Mattermost: Daily Standups and Peer Interaction

We use a #daily-standup channel on Mattermost to keep each other in the loop asynchronously.

You’re expected to post brief daily updates using a lightweight format like:

• What you did yesterday
• What you’re doing today
• Any blockers or questions

This isn’t for micromanagement. It’s to help you build momentum and self-accountability, share challenges transparently, and stay connected with your peers (who might have similar issues or ideas)

If you’re stuck or unsure, don’t hesitate to tag me or others—you’re not expected to solve everything alone.

We’ll also have a dedicated #masters-thesis-<year> channel for broader discussion and coordination outside of the daily updates.

Channels prefixed with #rp- denote research papers organized by topics.

---

GitLab: Organizing Research Tasks and Milestones

While your code lives on GitHub, we use GitLab to plan and track your research process. GitLab integrates natively with Mattermost, so project updates, task discussions, and check-ins can flow naturally into our daily standup or topic-specific channels.

Each student gets a GitLab project board with:

• A backlog of ideas and longer-term goals
• A weekly sprint column for tasks you’re actively working on
• A done column to track completed work (useful for write-up later!)

I suggest you update this board once a week, typically after your group check-in or as part of a Friday reflection. You can also use it to break down big goals into manageable steps, add links to code, papers, or Mattermost discussions, and track iterations of experiments and decisions over time

This makes your research journey legible—to both of us—and gives you a structured way to reflect and adapt as things evolve.

---

GitHub: Clean Code, Clear Structure, and Reproducibility

All project code should be version-controlled in your designated GitHub repository (under the Tapri Lab organization).

Why GitHub? It’s widely used and familiar, easy to share if we publish the work, and CI tools and archival workflows are simpler here

I suggest the following practices, especially since ‘research code’ is a bit of an unnecessarily derogatory term:

• Keep code clean and readable; think of future you, or a reviewer!
• Use commits freely, they are cheap. And write good commit messages
• Organize code, data, and results meaningfully (a simple README.md goes a long way)
• Push updates regularly, not just before deadlines
• Keep a dedicated notebook to generate each figure so it can be easily edited or regenerated later. Time spent on aesthetically compelling visuals is time well spent.

This isn’t about perfection. It’s about building good habits of reproducibility, visibility, and collaboration.

---

Weekly Rhythm: A Minimal-but-Effective Cycle

Here’s how all the pieces fit together:

• Daily (Mattermost)
  Post short updates in the daily standup channel to track momentum and ask quick questions.

• Weekly (Group Session)
  We meet for ~1.5 hours each week. While only 4-6 students present in a given session, everyone is expected to attend, listen actively, and contribute feedback. Presenters can give a short progress update, share something they’re working on, or walk through an experiment or idea. Peer discussion is strongly encouraged—it’s a chance to learn from others’ thought processes, not just their results.

• Weekly (Written Reflection)
  Writing is not just about recording results. It’s a powerful way to clarify your thinking. As fantasy author Brandon Sanderson often reminds his students, writing is good for you.

  Each week, I recommend capturing a brief reflection: What did you try? What worked or didn’t? What do you want feedback on?

  This could be added as a short note in GitLab, a paragraph in an Overleaf log, or even a scratchpad in your repo. You don’t need to write beautifully—you just need to write something. It helps us both keep track of your thought process, and makes your final thesis easier to assemble.

• Weekly (GitLab Update)
  After your session, update your GitLab board to reflect what you plan to work on next and what’s done. Use this to guide your upcoming week.

• As Needed (GitHub Commits)
  Code, experiments, results—commit and push regularly. Treat it as your source of truth.

Structured Embedding

This setup is not just logistical, it’s designed to help you feel embedded in the lab. Too often, Master’s students end up working in isolation, which can be both demotivating and limiting.

My goal is to foster group dynamics, collaborative energy, and cross-pollination with PhDs and other lab members. Later, you’ll develop your own style, and that’s as it should be. But this shared structure can help make the early stages less overwhelming and more connected. You’ll get more out of your thesis if you actively engage with the people around you, not just the problem in front of you.

Research, in my view, is at it’s core a people activity.