Thread by @emgurevich

How we are hiring algorithmic traders at @wintermute_t [Thread]

0) We’ve just completed 8 trader assessment batches, about 200 pre-qualified candidates completed screening tests, 40 people did our assessment center and 2 got hired. I'd like to share our process in hope that it is useful to others recruiting top talent and to future candidates

1) Hiring is mission critical for us and hence it’s done by the management team and the hiring team, not by recruiters or administrators (well, we don’t have these roles at all 🤷🏻‍♀️)

2) Our talent philosophy is “grow your own”, so we mostly hire traders at graduate or early career levels and we only promote from within (this is a potential subject for another post)

3) I personally write all job descriptions, screen, book meetings and reply to every applicant

4) I personally commit to “reply to all” promise and the depth of my reply is proportional to the effort *I think* the candidate put in the application (with the increased applicants pool keeping this promise was very tough, but I am still standing)

5) The first stage is pre-screening. Here we are looking for three things: 1) evidence of quantitative skills, 2) evidence of coding skills (ideally Python), and 3) motivation, evidence of interest in this role and in Wintermute

6) Motivation is the hardest to judge and verify hence I’ve built a few tricks in the application form (jobs.lever.co/wintermute-tra…)

7) “Have you read the job description?” Of course, no one says “No” but it nudges people to read if they didn't

8) “Why are you interested in @wintermute_t and in this role? Give a concise and authentic answer”. The answers are incredibly telling and help screen candidate further

9) While there is no right answer, there are definitely wrong answers, the most popular being: no answer at all, answer to a different question, generic/bs answer, answer that immediately shows culture mismatch

10) It also tests if people “listen well” (note the question asks for four things) and it further tests if they actually read the JD

11) If I see evidence of the three things above, I’ll ask candidates to do screening tests: one quantitative and one coding

12) If I see evidence of the three things above, I’ll ask candidates to do screening tests: one quantitative and one coding. It is meant to be quite easy to get to this stage

13) Quant tests is not advanced statistics, it's high school maths/problem solving. Coding is Codility test selected and adjusted by our traders that focuses on algorithmic tasks

14) Candidate who scored best are invited to the assessment center; in rare cases we do manual code reviews and put exceptional profiles though into assessment even if they scored a bit lower on one of the tests

15) Assessment center (AC, also called Round 2) consists of 1) two mental maths tests, and 2) a set of four 25-minute interviews with traders and developers

16) Unfortunately, only one AC was in person, at the time we also did a fun group market making game, cooked barbecue and had drinks with candidates. Now we have to resort to Zoom (hopefully not for long)

17) Why do we test mental maths? I guess part of it comes from traditional market making/hft testing legacy and, yes, modern traders don’t need to do mental maths all day long, but

18) It does test an ability to think under time and consequences pressure which is important for traders. Also, it’s a known fact in hft/mm that mental maths is going to be tested, so it implicitly tests for interest in the industry and..

19) It’s actually trainable. I know because I trained @EvgenyGaevoy for these tests 15 years ago and I trained myself for consulting interviews 5 years ago

20) Trader interviews cover a range of maths, probability, coding and experience questions. These are much harder to prepare for and we don't encourage it

21) The best candidates sincerely enjoy these puzzles (never hit-and-miss brain teasers!) and one one can figure out answers from general logic / first principles

22) We are big believers in systematic interview process, which means that we ask same questions to all candidates and compare performance across a large pool of respondents

23) One of the interviews requires preparation and I send out the assignment 24 hours before the interview. It’s not much time and that’s on purpose.

24) Just like in real trading environment, sometimes you need to do things on short notice without warning and you just need to find a way to squeeze it in

25) Unless interviewers are asking experience questions, they don’t see candidate’s CVs and credentials; this is meant to eliminate bias

26) Once interviewers interview candidates they put feedback in designated forms in Lever; they do not talk to each other, again to try to eliminate bias and group think

27) After all feedback is filled in (usually same day), we sit down together to review candidates. This is a structured and moderated session similar to performance review sessions; the decisions are wholistic, meaning we review all information we’ve got so far

28) The selected candidates are then invited to Round 3, which is interviews with management, typically head of trading, myself and CEO @EvgenyGaevoy (he interviewed every hire at @wintermute_t and every later stage candidate)

29) Management interviews are not about hard skills anymore, these focus on motivation and culture fit. Again, there are no good answers but there are wrong ones. People may want to join as for very different reasons but we should have trust in the authenticity of these reasons

30) The following would be likely “pass”es at this stage: solely financial motivation/mercenary mentality, lack of understanding of how we are different from competitors (incl traditional mm and quant firms), no potential to embrace “trader mentality”, high ego/not team player

31) After the offer is made, we typically set up a call where I explain in detail how our compensation works (bonus, equity) and offer to speak with anyone else at the company; it is a really two-way process at this stage

31) On the compensation topic, if someone isn’t willing to go through the process without knowing how much they will be paid, that’s screens out people who are not a match/ not motivated enough

32) Realizing that candidates invested a lot of time in the process, I find it very important to give candid feedback, esp. after Round 2 (before that the reason is typically test performance or not reading JD)

33) I've learned the value of feedback from my management consulting interview days where the hiring manager would call you same day and give detailed feedback. I remember appreciating that back then

34) I am doing my best to make it as candid and actionable as I can, and this is the part which I enjoy the least: telling “No”s to strong and motivated candidates who just didn't perform as well as others and finding way to give feedback when telling the "truth" is very hard

35) I also realize that no process is perfect and we can always be wrong about candidates no matter how hard we try

36) The best candidates I’ve seen are sincerely grateful for feedback and some even reapplied later having worked on improving certain skills (and some got an offer!)

37) That’s it! Let me know if you found it useful. Also curious to hear how different this process is from other trading firms in crypto.

P.S. I’d like to clarify that for different trading roles (e.g. DeFi trading), this process would be different. It would also be different for dev roles, operational and specialist roles, etc. The process should serve the purpose 😅