Advice for internship applications

The projects I supervise are:

• basic research: You will be asked to understand (and help formulate) an original research question, read the scientific literature, collect and analyze new data, and write about your findings. I will guide you through the process, but there will be uncertainty, setbacks, and the project will change over time.

• primarily computational: You will spend your time writing code and reports, running simulations, reading papers, analyzing data, debugging things that are supposed to work but decided not to, and generally trying to understand noisy molecular systems. You will be doing all that in front of a computer screen.

Before applying, make sure this actually corresponds to what you want to do, and that it matches the requirements of your program. You should have a look at my publications; there is no need to read everything in details, but that will show you the kind of questions I am interested in and the methods I use to tackle them.

Who should apply ?

Computational biophysics sits at the interface of several disciplines, and there is no single “right” profile to enter the field. I welcome applicants with backgrounds in biology, chemistry, physics, mathematics, machine learning, bioinformatics, etc.

Most importantly: this is research. Background and technical skills are important, but only insofar as you are genuinely curious and excited by the prospect of doing novel research in computational molecular science. The students who do well are usually those who:

• are genuinely interested in research and scientific questions
• enjoy understanding how and why things work
• tolerate uncertainty
• take initiatives
• are comfortable with abstraction and mathematical formalism
• do not get discouraged by mistakes or failed attempts

Research is fun because it is exploratory; but this is also what makes it difficult, slow, sometimes confusing and frustrating. My role as a supervisor is to help you learn how to do research, and hopefully enjoy it, without unnecessary pressure. Honest mistakes, confusion, lack of knowledge, and failed ideas are normal and expected. Lack of engagement, however, is much harder to work with.

If this sounds like a challenge you are excited to tackle, you are probably in the right place ! Conversely, if you are mainly looking for a short internship unrelated to research, or have little interest in computational molecular science itself, you are unlikely to enjoy working with me and you should probably apply elsewhere.

Expected background skills

Typically, my research projects involve a combination of:

• Biochemistry/structural biology
• Physical chemistry and physics, especially statistical mechanics
• Machine learning and numerical methods

You are not expected to know all these fields, but you should usually have at least some familiarity with two of them, and be willing to learn the third (How much familiarity depends on your academic level — see below).

A rough checklist

Here are examples of questions relevant for the research I do. They are not formal requirements, but they give a reasonable idea of the expected background. During interviews, I may ask questions like these to understand how you think and what concepts you are already comfortable with.

• Biology / biochemistry
  • Cite a few examples of proteins and describe their biological roles
  • Describe (verbally) what an $\alpha$-helix or a $\beta$-sheet looks like
  • Explain the difference between folded and intrinsically disordered proteins
  • Explain what ATP is and its main cellular role
• Physical chemistry and physics
  • Write Newton’s second law and explain what it means physically
  • Explain qualitatively what temperature represents in statistical mechanics
  • Write the Boltzmann distribution for a classical one-dimensional harmonic oscillator
  • Explain qualitatively what a free-energy landscape represents
• Machine learning and mathematics
  • Explain the purpose and principle of principal component analysis (PCA)
  • Solve a first-order linear differential equation such as $y’ + ay = b$
  • Explain the difference between training and testing data in machine learning
  • Explain what an activation function is and why it is needed in deep learning

If basically nothing on this list is familiar to you, then the research I do and the projects I supervise are probably not a good match for your current background.

Expected technical / programming skills

Because the projects are computational, some amount of technical proficiency is necessary in addition to broader disciplinary knowledge. Useful skills include:

• Linux command-line usage
• Basic Python programming
• Scientific Python libraries (Numpy, matplotlib, Jupyter notebooks, etc)
• Molecular visualization/simulation softwares (Pymol, VMD, ChimeraX, Avogadro, GROMACS, NAMD, OpenMM, etc)
• Data analysis & Machine learning libraries (pandas, scikit-learn, pytorch, etc)

Prior experience with Molecular Dynamics simulations is helpful, but not mandatory for junior students. In any case, I care much more about whether somebody is capable of learning technical material than whether they already know a specific software package. Most technical skills can be acquired reasonably quickly if the underlying scientific motivation is there.

Types of Internships and Expectations by Academic Level

My expectations naturally increase with academic level. An M2 student applying for a six-month research internship is not evaluated by the same standards as an L3 student discovering research for the first time.

L3 / M1 Students (final-year Bachelor’s or first-year Master’s level)

For L3/M1, I mainly look for scientific curiosity, seriousness, and the ability to learn progressively with guidance.

Typical internship:

• Project scope: small scale, self-contained research project taylored to your skill level
• Duration: 2-3 months; M1 students can apply for longer stays (up to 6 months)
• Compensation: no compensation if the internship lasts less than 3 months, ~600€/month otherwise

Projects at this stage are designed primarily as research training. The main objective is to discover how computational research is actually done: reading papers, handling data, running and analyzing simulations, discussing results, and progressively learning how to think about scientific problems in a more autonomous way. Publications and conference presentations can result, but are not particularly expected.

Expected background

At this level, I usually expect:

• solid undergraduate-level foundations in at least one relevant scientific discipline
• basic programming and data-analysis skills
• enough scientific maturity to read scientific papers with guidance Advanced technical expertise is absolutely not expected at this stage.

Example projects

• Analysis of peptide folding simulations
• Coarse-grained simulations of intrinsically disordered proteins
• Machine-learning analysis of molecular conformational ensembles

M2 students (second-year Master’s level)

At this stage, students are expected to function as junior researchers rather than as students discovering the field for the first time. Accordingly, M2 internships are substantially more demanding and are usually designed to lead-up to a PhD project (but you do not have to commit to staying for a PhD right away).

Typical internship

• Project scope: a relatively autonomous and self-contained research project, often laying the groundwork for a PhD topic
• Duration: typically 6 months
• Compensation: standard French M2 internship stipend (~600€/month)

The level of independence expected is correspondingly higher. I do not expect students to know everything already, but I do expect them to progressively become capable of identifying problems, reading documentation and papers independently, debugging their own workflows, and presenting/discussing scientific results critically.

Expected background

At this level, applicants are generally expected to have:

• reasonably strong programming skills (Python preferred)
• familiarity with Linux and scientific-computing workflows,
• prior exposure to molecular simulations, statistical mechanics, machine learning, or related computational methods
• the ability to work mostly independently after initial guidance

Example projects

• Enhanced-sampling molecular dynamics of gas-phase peptide oligomerization

During the Internship

Environment

You will have your own desk and workstation (with Ubuntu), and access to supercomputing resources. I am part of an interdisciplinary research group that mixes computational science with wet lab research in biochemistry and structural biology. So, you will have multiple opportunities to interact with other scientists including experimentalists. Presence hours are flexible as long as progress is satisfactory.

Supervision

I will personally supervise your work, possibly with assistance from an experienced student (M2/PhD student) or post-doctoral researcher. I am available for technical help; I expect you to seriously try on your own, but also not to spend more than a couple of days stuck on a technical problem, especially at the beginning of the project. We will have one-on-one ~1h update meetings every two weeks. Update meetings can be more frequent if requested, or conversely can be dropped down to once a month if after a few weeks it is clear that you can stay on track. There are also semi-regular group meetings, where you are expected to present your results in a few slides.

Scientific writing

For your work to be useful to other scientists (and even to future you), it is critical that you write about it and how you did it. In addition, I believe very strongly that writing about one’s research is the best way to clarify one’s ideas and scientific thinking. And, professional scientists are largely judged on the quantity and quality of their scientific writing (publications, research proposals). Scientific writing is therefore a crucial skill, yet it is often overlooked in academic training programs, especially before the PhD. For these reasons, I place a particular emphasis on writing for all my students.

To promote the habit of regular scientific writing, I will expect monthly reports and encourage you to start writing your official report/thesis on week 1. Monthly written progress reports (1-4 pages) should concisely describe the work done, the main findings (with figures), their discussion, and the next steps.

Once you have a sufficiently advanced draft of whatever important piece of writing you are supposed to deliver (thesis, research paper, etc), we will have interactive feedback sessions where we sit down and review your draft word by word together. These sessions are intense, but they are an exceptional opportunity for your supervisor to help you refine your writing and thinking up to academic standards. They will make you grow as a scientist and you should actively seek them.

AI Chatbots

I urge you to use LLMs sparingly and never as a substitute for personal reflection. I am aware this technology and its uses are evolving rapidly, which makes trying to edict strict rules pointless. Still, use your best, honest judgment.

Here are examples of what I consider acceptable uses of an LLM:

• Getting acquainted with a new field or concept
• Generating simple scripts for basic tasks (things like using bash to rename a large amount of files according to some non-trivial pattern, or tweaking the layout of a figure in matplotlib)
• Improving a script/code you have written yourself
• improve the flow and clarity of an already structured piece of writing

By contrast, I am cautious about the following use cases:

• Turning a bullet list into a structured text, if it is intended for a piece of potentially published scientific writing (e.g. your thesis or a paper).
• Generating critical code from scratch (a.k.a. vibe-coding your project)
• Generating illustrating images
• Using an LLM as the sole source for a detailed literature search

In brief, always be mindful and critical of the output and make sure you understand it. I will ask you to explain specific parts of your code or your writing ! Finally, keep in mind that feedback will help you improve only if it is your work you are showing. Thus, relying on an LLM would just deprive yourself of an opportunity to improve crucial research skills, and generally waste your time and mine. I have reviewed enough AI slop internship reports to attest of that.