Preamble: IP Paris has modern rules on how to obtain the HDR. They aim at rigorously ensuring the quality of the HDR degree from this institution, but at the same time making this process efficient and omitting many unnecessary excercises. On this page, I collect some hints on how to successfully navigate this process. They do not repeat the full content of the official rules, so please read them as well. Needless to say, the official rules are what counts in case of doubt (note also the dates of the conseil académique at the bottom of the page). Here is my copy of these rules (but you will also need the forms from the official page at some time).
This page is very recent. If you find a mistake, a part that is hard to understand, or you feel that something critical is missing, please contact me (Benjamin Doerr).
Recommendation: Since the way to this diploma takes some time, even from the point on when you are ripe for it, I strongly suggest that you start the HDR procedure as early as possible. There is no reason to wait longer than necessary. You only risk that at some time you want to supervise a PhD student and then you cannot do so (yes, in theory there is a derogation process, in practice you do not get this done in time).
Hence important ingredients of a successful dossier are publications in recognized international journals and successful (co-)supervisions of M1 or M2 internships or PhD students (different from other places, we do not require that you have supervised a PhD student as main supervisor via a derogation and we do not want this - if you are able to supervise, simply get your HDR). Strong conference publications are important as well (depending on the customs of the area), but you still need journal publications. Anything else that shows your recognition in the community is helpful, but not essential (e.g., memberships in program committees or juries for prizes). Third-party funding is only helpful if it really shows that you are a great researcher and definitely cannot be used as excuse for lacking publications. Teaching does not help for the HDR, but it is a good idea to mention it in your dossier as it helps understanding it.
Due to the very different customs of the different research communities and due to the different ways a dossier can be convincing, it is impossible to give a simple rule what is enough. Maybe five journal publications and two supervisions are a reasonable first guess, but it really depends strongly on the details. Publications in fake journal are strong minus, same as anything that suggests that you have not understood what is good research or supervision. Time is irrelevant apart from the rule that you should have five years of research experience after the PhD.
If you are not sure whether you are ripe for the HDR, simply send me - in my role as référent HDR of the IDIA (computer science, data, and AI) department of IP Paris - a CV that in particular covers the points discussed above. Statistically speaking, most people start their HDR process later than necessary.
My very strong recommendation is that you send this dossier first to me so that I can check if it looks good. You do me (and everyone else) a favor if you send the forms simply as word document (with you signature in). This appears silly, but the world does not become better if we all print each document, sign it, and scan it again.
While the official documents do not require so, we ask you to also submit the names of your referees with your authorization request (this is one more form). You need (i) at least three in total, (ii) at least two from outside IP Paris, (iii) at least two with an HDR (or equivalent), (iv) at least one international one. Needless to say, they have to appear competent in judging whether you deserve the HDR, in particular, by reading your thesis. The more competent, the better. Again, a good idea is to discuss this with me beforehand. My recommendation is to have all three reviewers with HDR or equivalent to avoid a possible bug in the rules (the rapporteurs will be part of the jury and all jury members need to have the HDR or equivalent).
You can submit your dossier anytime you want. The scheduling bottleness is the conseil academique, who meets roughly every two months. You find the dates at some unlinkable location of this page. If you submit your dossier one week ahead, you have good chances that it will be treated at the next CA. If I haven't seen your dossier before, it might take longer.
The research synthesis is usually between 20 and 40 pages long. Longer formats are accepted when there is a particular motivation. This should be discussed beforehand with the referent HDR. In no case, there should be longer passages of text identical to ones in scientific works published by the candidate. Also, no original works should be appended to the research synthesis.
You submit your thesis to Emmanuel Fullenwarth (cc to me). You will have to send it to your rapporteurs yourself. Discuss with your rapporteurs how long they will need to review it. This cannot be less than a months. Between the reports coming in and the defense you need again a good month, discuss the timing with Emmanuel Fullenwarth.
The jury consists of at least 5 members (including the reviewers), all with HDR or equivalent, at least 50% outside IP Paris, and at least 50% rang A (professeur, directeur de recherche, etc.). There are no rules preventing the jury from containing people that appear close to the candidate (but the jury has to be validated by me, the conseil académique, and the president of IP Paris). My recommendation is to have around eight jury members and to choose them as competent, shiny, and independent as possible. That said, one could think of having one jury member from the lab (but maybe not the same team) or at least IP Paris to show that the lab or IP Paris had their say on the result. One can also have the own PhD supervisor in the jury. In some areas this is common (and an offence if you deviate from this custom), in others not (and seen as strange since this person most likely is not fully unbiased).
If you are sufficiently close to being ripe for the HDR, then you can ask for a single derogation for a single, specified PhD project. This makes a lot of sense in universities where such a PhD supervision is required for the HDR. Since we do not have this rule, there is less need for this type of derogation. I note that such a derogation is not much easier to obtain than the HDR (you need all the avis favorables plus an extra one from the ecole doctoral). Also, since you need such a derogation specific for a particular PhD project, the timing is non-trivial (you need to know that you want to conduct this project, typically because you have a good student for it, and then you need to get the derogation, get it into ADUM, get the project in ADUM, get the grant for the student, and get the student registered). In consequence, at least 95% of the people who ask me for this derogation fail to obtain it in time and then have to resort to other, suboptimal measures. The majority of these could have easily obtained their HDR had they started it two years before. Long story short: Just do your HDR in time and save you from the hassle to obtain this derogation.
[1] Five minutes is the estimated reading time for this document. Depending on your circumstances, further and more time-consuming steps might be necessary. Nevertheless, the HDR procedure at IP Paris is designed to be an efficient pathway to the HDR for those who satisfy the high standards of this institution, so there is no reason to hesitate.