In 2015-2016, I applied to permanent computer science research and teaching positions in France. As I did so, I was surprised to see that, in most circumstances, there was very little public information available about the application process, and important clues had to be gathered by word of mouth, rumors, and indirect questions to jury members. I found this a bit problematic at the time, and thought it would be helpful if former candidates would post their application material online, so that other candidates can know where to start and what to expect.

As I am done with applications for now, this page is my attempt to make this happen. It presents the various applications I did, the documents that I submitted (with some details anonymized), some pointers and notes and contextual information, and answers to questions which I obtained at the time. The page is written in English to help international candidates, but most of the documents are in French.

Disclaimer: This page only reflects my personal opinion. It may become outdated as the rules of these competitions evolve. I am certainly not the most qualified or authoritative person to tell you about these topics, I just thought that it could be useful to someone to produce a brain dump of my experience with them. No information posted on this page has any official status and none of it was approved by any of the institutions mentioned. Don't forget to use your best judgement.

Also, this page is not intended as a serious criticism of how these competitions are run. Quite frankly, I think the main problem about them lies somewhere else, namely: the overall number of available researcher positions is ridiculously insufficient. More details at the end.

CNRS

I applied to a CNRS CR2 position in section 6. I was preselected but I was not ranked in the admissibility list.

CNRS is France's largest governmental research organisation. Among its various job application mechanisms (including vacancies for administrative staff, research engineers, etc.), it organizes a yearly competitive examination for researcher positions at various levels. These highly coveted positions carry no teaching obligation and (traditionally) little administrative oversight, except that of handing in a regular activity report roughly every 2.5 years (details here). The salaries of CNRS are fixed by law: they depend on the level and your former career, and include various bonuses (Prime d'encadrement doctoral et de recherche, Prime d'excellence scientifique, etc.): as of 2016 a new researcher at the lowest pay grade can hope for around 2000 EUR/month after tax. CNRS positions are fonctionnaire positions, meaning that CNRS researchers are directly employed by the French state. As of 2016, the competition formally opens on December 1st, and closes in early January, although of course applicants who are in the know will usually get started in October-November.

At the right time of year, the offers are posted here. They are sorted by "Section": the sections most relevant to computer scientists are 6 (essentially: fundamental computer science and some other things) and 7 (signal processing, image processing, language processing, robotics, and other things); the boundary between the two is often blurry. You can apply to both if it makes sense.

The contests are then ranked depending on the kind of job. From most senior to least senior, they are "DR2" (directeur de recherche de 2e classe), "CR1" (chargé de recherche de 1re classe), "CR2" (chargé de recherche de 2e classe). As per section 6's recommendations, CR2 positions are typically intended for people with 7 years or less of research activity, counted since the beginning of their PhD (and excluding career interruptions, e.g., having a child). There may be some special contests with a position that comes with some constraints, either thematic or geographic. Independently from this, some positions of a contest may be assigned "in priority" to some themes, although these recommendations are not always followed. In practice, your application should target one or more of these contests (e.g., "06/03" is a contest): if you fit one of the themes you can mention it in your application, but it's the same application no matter the theme(s) you are focusing (if any).

The application process is entirely done online. The only official source of informations about the application is the CNRS website. In practice, however, you should also look at the non-official information posted by the juries of the sections to which you apply, e.g., the current websites of section 6 and section 7. These websites often contain important information from the section to the candidates, and they also present the jury members: it is a good idea to look at the jury members, to try to guess which ones will review your file, and make sure that it is framed in a way which will suit them. (People in the jury whom you know personally or with whom you have worked will usually declare themselves in a conflict of interest and will not participate to the selection process for you.)

When submitting your file, you will indicate to which laboratories and teams you will like to be assigned. Indeed, CNRS researchers are commonly assigned to research teams in universities that have a joint lab with CNRS, in what is called an unité mixte de recherche. You will need the UMR number of your host laboratories when applying: search for it online or ask the teams for help. The recommendations of section 6 indicate you should have at least two proposed host laboratories. Rumor has it that it is essentially impossible to apply to a host laboratory where you did your PhD, or where your PhD advisor has moved (though there have been some exceptions). In any case, it is common knowledge that you should already be in touch with your proposed host teams when submitting your online application. Usually, well before December, candidates will meet the members of their target teams and discuss their application, and laboratories will decide on which candidates to support. It is essentially useless to request affiliation to a laboratory or team if the team and laboratory do not already know you and support you, because it would put you at a disadvantage compared to candidates who have this support.

For the first phase of the process, i.e., the online applications, here are the required documents:

It is customary for candidates to request some feedback from their potential host teams about their application file, in particular the project and how they would integrate to the team. It is also customary for candidates to ask former candidates whom they know, so that they can have more examples of application documents. Candidates can choose to disclose their order of preference among their requested host laboratories, or not; of course it is unavoidable that there is usually a fair amount of bluffing between all involved parties at this stage...

For the second phase of the process, a preselection list is established by each section. All candidates on this list proceed to the next phase, the others are eliminated for that section. This usually narrows down the pool, e.g., to around 20-30 applicants for CR2. (Historically, this phase did not exist, so the section had to interview all applicants, including those that they did not want to consider. Now that sections can pre-select applications, they are using this power more and more, so the historical trend has been for the number of preselected people to decrease over time.)

The third phase of the process is an audition, for all preselected candidates. The week of the auditions should be publicized sufficiently in advance for you to make travel plans (although it is common to hear the information in advance from your favorite non-official sources...). Later, a schedule is established and you should receive an official appointment. The section should indicate information about how the auditions will proceed: for section 6, it included a presentation for 15 minutes, then questions for 15 minutes, but this often changes from one year to the other (expect instructions in your "convocation" or from non-official sources again). For the audition, you will need:

In 2016, these documents had to be on a USB key, to be brought with you on the audition. (Of course, bring several, in case there is a hardware problem.) No dress code was specified: I wore a suit, but most of the candidates I saw did not.

The audition is not supposed to be the main criterion on which the jury will rank people: in principle, the paper file should not be entirely disregarded once you get to the audition stage. Of course, in practice, your live performance probably has a significant impact. As the CNRS contest is ridiculously competitive, I think it is safe to say that you need an extremely good file and extremely good audition and extremely good fortune to get in, unless maybe if you trump the others on some criterion (think settling P vs NP).

At the end of the auditions, an admissibility ranking is issued by the jury of each section. The ranking is a sorted list of some subset of the preselected people, and it may include ties; the number of ranked candidates is usually greater than the number of positions. Anyone not on this list is eliminated for that section (this was my case), everyone on the list proceeds to the next phase.

The fourth phase of the process is that, in mid-June, final admission rankings are issued. This is where it gets complicated: in addition to the sections, CNRS is split into institutes. Sections 6 and 7 are attached to the Institut des sciences de l'information et de leurs interactions (INS2I), and it is this institute which establishes the final admission list, by picking a subset of the admissibility list and some order among this list. (It is not exactly clear whether the order on the subset may be something else than the restriction of the original order to the subset, i.e., if the INS2I has more options than merely eliminating some candidates.) In particular, the admission list may even be shorter than the number of available positions (in which case some of them will not be staffed). The tradition used to be that the INS2I would not change the lists established by the sections, but in 2016 the lists were substantially changed for complicated reasons, so candidates should now expect possible surprises at this stage.

Once the final admission ranking is established, an offer is extended to candidates, in the order of the list. The question of choosing the host team of accepted candidates is usually settled through informal discussion between candidates and the CNRS administration: it is not unheard of that the location preferences of a candidate would have an impact on their final ranking.

Inria

I applied to Inria as CR2 in the centers of Lille, Saclay, and Sophia (actually in Montpellier). I was preselected at Lille and Saclay, ranked second in Lille and fourth in Saclay on the admissibility lists, and was ranked in the top-2 at Lille in the admission list.

Inria is a French research institute which specializes in computer science. It also organizes a yearly competitive examination for researcher positions at various levels. The rankings (CR2, CR1, DR2) are the same as for CNRS above, Inria researchers are also employed by the state (fonctionnaire), and the pay grade is similar to CNRS (with some minor possible differences). Inria positions are also pure research positions with no teaching obligation. However, the cultures of Inria and CNRS differ in some respects:

As of 2016, the Inria competition formally opens in January, and closes in mid-February, although of course applicants who are in the know will usually get started earlier. The positions are assigned to each centers; there are usually one or two CR2 positions available in each center. The non-Parisian centers are rumored to be less selective, but of course it depends on who is applying where.

The application process of Inria, surprisingly enough, has nothing to do with CNRS. A Word/LaTeX template is provided for you to fill, with a mandatory structure that you must follow. So there are less doubts about how it should be done, but you have little freedom to evade questions about things that you would rather not talk about (e.g., software development, or industrial collaborations). Another difference is the length: the summary of your past work is essentially limited to a one-page summary and three one-page "contribution sheets", and your research project is limited to 2 pages (again, its contents are not binding). Further, Inria applications must be sent by registered paper mail, not submitted online, although in addition to the paper applications you should also email an electronic copy of the application document to an email address which is indicated. Sending the paper application is mandatory and it has happened that candidates were disqualified because of issues when using postal mail. Last, as Inria is organized in regional centers, you apply to each center separately, and send a separate document for each: you can have, e.g., a different research project for each center; but you can also re-use the same project and only change the explanation of why it would integrate well with the team, as I did. It's probably better to have different projects if it is required to fit better to the themes of the various teams where you apply, and to possible collaborations with its members; the only downside to having different projects for each team is that it takes more time for you to prepare the application files.

As some laboratories are joint between universities, Inria, and CNRS, it is entirely possible that you will find yourself applying to the same team through three independent processes: Inria, CNRS, and maître de conférences (see the corresponding section below).

For the first phase of the process, i.e., the online applications, here are the documents I submitted. It was possible to attach a photo, but I did not do so.

If I remember correctly, an evaluation committee in early March at the national level decides on which files are eligible to compete. Again I had the problem that I had not defended my PhD yet, but it is possible for this committee to accept an application even without a PhD. Candidates which are not kept at this stage are eliminated altogether.

The second phase of the process is that each Inria center decides on which candidates are preselected. Candidates who are not preselected for a center are eliminated for that center; candidates who are preselected for a center will have an audition in this center. The planning for the preselection list and for the audition periods is different for each center, but auditions in 2016 were in April-May. (As for CNRS, you will probably want to find out quickly about this, to arrange your travel plans.)

The third stage consists of the auditions. They usually include a presentation and questions, similarly to CNRS (see above): the duration of each depends on the center (usually it's around half an hour total). For some centers, you will have to email your presentation in advance. I was auditioned at Lille and Saclay, so here are:

The (lack of) dress code was the same as for CNRS.

Shortly after the auditions, each Inria center establishes an admissibility ranking. Candidates not in the list for a center are eliminated for this center.

The fourth phase of the process is that, in mid-June 9roughly at the same time as for CNRS), a final admission ranking is published. There is one ranking per center, but these rankings are all decided nationally, and the lists usually vary wildly between the admissibility and admission rankings. In particular, the committee usually modifies the list to ensure that each candidate is only admitted to one center, so they can decide who they want to send where. Many candidates are routinely eliminated at this stage for inscrutable reasons, no matter their admissibility ranking. Admissible candidates are usually informally asked about their preferences, to be given as input to this mysterious process. (I was asked for such preferences.)

Offers are then extended from the various centers to the candidates in descending order of the admission list for this center. Positions include a salary estimation, and candidates usually have 2 weeks to accept or decline them.

Maître de conférences

Positions of maître de conférences (tenured associate professor) are half-research, half-teaching (with a reference amount of 192 hours/year of teaching in "équivalent TD"), with a pay grade fixed by law and similar to CNRS/Inria above (including some primes) and a fonctionnaire status. They require a diploma known as the qualification (more on this below). They usually open in March as part of the session synchronisée. There are exceptions, however: I applied to Télécom ParisTech and none of the previous points in italics were true! Other well-known exceptions include, e.g., the École polytechnique.

The normal process is that you apply for the qualification in November-December, and then apply to positions on the Galaxie system. However, the specific application process differs among the universities. I will not talk about this as I have not done it.

The position to which I applied in Télécom was posted a bit later than the session synchronisée, on Télécom's system, in May. I provided the following documents:

I was then auditioned in an informal interview with the team:

There was a shortlisting, and then I had a formal interview:

A ranking was established, and the recruitment was confirmed by the director of Télécom.

Parting words

Applying to research positions in France is a very competitive process. To some extent, this is healthy; but I do not think that the current state of affairs is reasonable. Candidates spend an excessive amount of effort to optimize their applications and to guess and follow the rules of the game, and most importantly the number of available positions is simply not sufficient to accommodate all candidates who would deserve to obtain one. The French system, after having spent significant amounts of money to educate promising future researchers, ends up discouraging too many of them, who will have to do jobs that they enjoy less, leaving lots of important research undone. This is especially infuriating when one sees, e.g., that the state allocates more than the budget of CNRS to a confusingly named Crédit d'impôt recherche, a wide-ranging tax exemption for companies, with only a tenuous connection to research, and which has attracted significant criticism.

This sorry situation should be kept in mind by applicants and interested parties, but of course it is also a sore loss for France in general.