My Journey from Econ PhD to Tech — Part 2: Interview experience
Links to other parts of the series:
Part 1: Interview prep + Networking
I know it sounds like cliche at this point, but my interview experience again showed how important connections are: at firms where the hiring manager picked up my profile to begin with, i.e. Uber and Zillow, I had an extremely fast and smooth process; at firms where I applied through the general pipeline, i.e. Facebook and Google, my experience was painful and long.
The Start
On Aug 7th, a Stanford econ phd circulated a Data Scientist opening at Uber in the Stanford econ phd mailing list, copying the hiring manager. Seeing it, I realized that this is the opportunity — it’s a very directed search, meaning that the hiring manager was probably looking for econ phds in particular; it is very early in the cycle, meaning that most people probably aren’t prepared yet, and if I apply I might be the only candidate, and if they are looking for people to start soon, I could have a good chance getting in (because I’m graduating early).
But at that time I weren’t 100% ready yet: I was still in the process of completing my EDS crash course on ML, and haven’t finished all 3 ML projects that I wanted to have on my CV. — It turns out that I was overthinking — in my entire job searching process, I was only asked ML questions in the Zillow interview where the job title literally has ‘Machine Learning’ in it. Moreover, they hired me not for my ML skills but for my econ phd and domain knowledge in housing (they are a team of ML specialist looking for an econ phd who knows how to think about the housing market).
Here’s another thing that the Laioffer class was useful for: in the first class the teacher told everyone that ‘You’ll never be ready ready — if you feel 100% confident when you applied, you probably started too late’. That made me realize my urge of wanting to finish all the ‘preparation’ was probably wrong, and I replied to that email 4 days later on Aug 11th Tue.
The next day, on Aug 12th Wed, the hiring manager replied to schedule a chat tomorrow, i.e. Aug 13th Thu. The chat went well — it wasn’t an interview, but just a casual chat to understand my background, how serious I was about the tech market, and my timeline. The hiring manager told me to apply via their career page, but also to let her know when I did so, so she could help escalate it to speed up the process. In fact, on Aug 17th Mon, after seeing that I still haven’t applied, she emailed me to check what was going on. — what was going on was that I was trying to finish doing all the ML projects I wanted to put on my CV… which as I said earlier turned out to be completely unnecessary. Upon receiving her email, I applied for the job right away — recruiting season has officially started for me!
Knowing how important aligning timeline is for wage bargaining purpose from reading Haseeb’s story and from the ‘job market candidates panel for tech’, I figured I really should start to apply for other jobs too. Thus, in the same week, I pinged all the contacts I collected during the networking process who had said they can refer me, to hand them my CV and tell them I’m ready to apply. It’s also in that week that Anthony posted on his Twitter to help me find more jobs (which turned out to be not so useful).
Uber
On the same day as I applied for the job, i.e. Aug 17th Mon, the recruiter reached out to me to schedule a chat. We chatted on Aug 20 Thu, where she told me about the structure of the data science team at Uber. The next day, Aug 21 Fri, I received confirmation for my phone screen, which was scheduled for Aug 26 Wed.
The first phone screen is always the hardest — I still had no idea what a tech data science interview looks like, and what they were looking for. Also, Uber’s interviews are heavily case study, which is the most fun, but can also be the hardest, because case studies are very open ended. I thought I did terribly, and went into a bet with Anthony on whether I passed, which I failed — on the same day, the recruiter called to say I passed, and that I’ll be invited to a second phone screen.
At that time, I thought what happened was that I did so badly that they decided to add a phone screen, which turns out to be completely wrong — firms don’t change their processes for individual candidates. I’ve indeed heard of cases when someone had 3 phone screens, but it’s usually because of something on the team/firm side, rather than because of the candidate.
The next day, Aug 27 Thu, I received confirmation for my second phone screen on Sept 3 Thu. — if I had asked for it to be scheduled sooner, I’m sure it would have happened; at that time I wanted it to be scheduled for later, because Uber was moving too fast compared to my progress with other firms.
On Sept 3 Thu I had my second phone screen, which went much better — the second one is basically the same as the first one (after the onsite I realized that all interviews at Uber, either phone screen or onsite, are basically the same, i.e. case studies). The interviewer even asked ‘have you done an internship before?’ (I don’t) because she was surprised by how practical I was.
On the same day, the recruiter called again to say I’ve passed the second phone screen, and collected my availability for the onsite. On Sept 4 Fri, I received confirmation for my onsite for Sept 11 Fri, i.e. 1 week later.
Notice how amazingly fast things were — if you don’t think it’s fast, don’t worry, you’ll realize it by the time you read my experience with Facebook and Google.
On Sept 8 Tue, I had a super detailed prep call with the recruiter on what would come up in the onsite — most firms do this with their onsite candidates, but firms differ on how detailed and how accurate the info are. Uber is on the detailed and accurate side — I’ve had cases where what the recruiter said is completely different from what I had.
The recruiter was on vacation from Sept 11 Fri, i.e. my onsite to Sept 14 Mon, so I only got to know my result on Sept 15 Tue: I got the offer! The first offer is always the most relieving because you know the rest is cherry on top. — which turns out to be not true, because having multiple offers is extremely important for wage bargaining. I highly recommend you not to end your recruiting journey after your first offer, even though all parts of you are screaming ‘I want to be done!’ at that time.
On the whole, the Uber recruiting experience was extremely smooth — the recruiters were super fast and responsive and know what they are talking about. The interviews were fun and I learnt a lot. All the interviewers were friendly. The hiring manager was extremely supportive — always there to answer my questions regarding the recruiting process.
The only thing I did not like is that they are on the more pushy side in terms of deadlines: when they told me my offer details (i.e. $$$) on Sept 18 Fri, they didn’t say there will be a deadline. But on Sept 24 Thu, the recruiter suddenly scheduled a call to tell me that Oct 9 Fri is the deadline. Now that I think about it, it’s probably because, in the meeting with the hiring manager’s manager on Sept 22 Tue, I said that ‘with high probability I’ll be signing’, and I told the hiring manager that the only other interview process I’m in at that time was with Amazon and my onsite will be on Oct 2 Fri.
This shows one of the most important principles when managing your recruiting process — don’t share too much information! Information = power. Whoever has more information has more power. In the recruiting process, the firm has the upper hand: you’re usually very young and naive, answers whatever question the firm asks such as ‘what’s your progress with other firms’, ‘what’s your salary expectation’, ‘what do you feel about the offer’, ‘when can you make a decision’, etc. You really need to protect your information, and get as much information from the firm as possible, e.g. ‘what’s your progress with other candidates’, ‘what’s the salary band for this role’, ‘how’s my interview performance — can you share the notes with me’.
Back to Uber — with a bunch of pushing, they not only canceled the Oct 9 deadline, but did not set a new deadline at all. Because of the Oct 9 deadline, I had to expedite my interview process with other firms by a lot, and eventually had 4 back-to-back onsites on Sept 30 Wed & Oct 1 Thu (Zillow, split into 2 days), Oct 2 Fri (Amazon), Oct 5 Mon (Facebook, Novi), Oct 6 Tue (Google). Back-to-back onsites aren’t a good idea — towards the end, I can’t even remember which examples I’ve used for my behavioral questions that day.
That’s why I’m telling you that all deadlines are fake, unless there literally is another candidate who will sign her/his name on the contract tomorrow if you don’t do so today, which is almost never the case. If a firm says ‘you have to decide within 5 days/1 week/2 weeks of the offer’, just say ‘This is a really big life decision. I’m afraid that this timeline won’t work for me at all. I’m still in the process with a few other firms I’m equally excited about, and I’d like to have all the information I need to make an informed decision. I don’t want to go back on my own words, so I don’t want to commit pre-maturely.’ — It costs a firm ~$25k and 45 days to arrive at the offer stage with a candidate, so once the offer is in your hand, you have the upper hand, and it’s them that don’t want to lose you.
Zillow
If you think the Uber timeline is amazingly fast, check out my experience with Zillow: Sept 24 Thu initial chat with recruiter, Sept 25 Fri phone screen, Sept 30 Wed & Oct 1 Thu onsite (split into 2 days), Oct 1 Thu offer.
As I mentioned in the previous section, I talked to XN, a Senior Applied Scientist at Zillow Offers in late Aug. Even though there weren’t any positions for my level at that time, he said he will let me know as soon as something opens up. For most people who had said something like that, they never followed up, but XN did: in mid Sept, he reached out to me to say a position that fits my profile will open up soon, and asked if I’m still available. I said yes, and he passed my CV to the hiring manager. On Sept 24, he said that the hiring manager has expressed interest. The rest is history.
The reason why Zillow was moving so fast was that I’ve let them know my Oct 9 Fri Uber deadline + the hiring manager was very interested in my profile (they are a team of ML specialists looking for an econ phd with domain knowledge in housing) + the recruiter was really awesome.
On the content front: the role that I was applying for is called ‘Applied Scientist, Machine Learning’, so not surprisingly, there are ML questions, but still presented in a case study fashion — they would present you a real problem (or a stylized version of that) that they were solving, and ask what you would do. It was super fun — probably tied with Uber in terms of fun-ness. The hardest coding question I had in my entire recruiting journey was probably also with Zillow (maybe tied with Facebook Novi). Even though I wasn’t able to solve the whole thing, the interviewer was happy that I got the spirit of it very quickly after he gave me a hint, which to me was also a positive signal — they care more about my problem solving ability, less about the technicalities.
Coursera
Remember that I said the same week when I applied for Uber, I also pinged a number of my contacts to say ‘I’m ready to apply, please refer me’? EN was one of them. I got to know EN in the department’s ‘job market candidates panel for tech’, and he generously referred me to all 3 firms he worked at before, and 2 of them turned into offers.
On Aug 20 Thu, EN wrote an email introducing me to EY, a very senior person on the Data Science org at Coursera. I was both surprised by how high up the connection was, and also by the fact that she offered to chat — we talked the next day, Aug 21 Fri. Next Mon, Aug 24, the recruiter reached out to collect my availability for the phone screen, which was later scheduled for Sept 2 Wed.
I thought I was interviewing for their full-time position, but at the end, the interviewer (who also turned out to be my hiring manager/manager) said, ‘You did great! I can offer you an internship right now, or we can re-connect next spring when we know our full-time headcounts for 2021.’ Lucky enough, I’ve already learnt the most important lesson for job hunting — always say yes, whether it’s a connection, or a referral, or an interview. I gladly took the internship offer, and started a few weeks later on Sept 21 Mon.
Even though the internship didn’t eventually convert into a full-time offer — during the midterm review, DN (my manager) said I did great and if they had headcounts they’d give me an offer, but they still don’t know their headcounts yet — the Coursera internship was still an extremely correct decision: practically speaking, I got to know what it actually looks like working as a Data Scientist in tech, and got experience with basic tools such as SQL and Python — interview prep and real work is still different: in interviews, my SQL code was 5 lines long; at Coursera, they were 50–500 lines long. But more importantly, I just had a really really good time — my team were awesome and I enjoyed talking to everyone — the ‘weekly hang’ was the thing I looked forward to the most every week during that two months.
Quora
Anthony not only posted on his Twitter about my job search, he also pinged his friends for referral, one of which is someone whom he got to know during his Facebook time who later worked at Quora, who then introduced me to his friend, WN, a Data Science manager at Quora. (This is similar to how I got to know JN at LinkedIn: Anthony introduced me to his friend GO, who later introduced me to him.)
Out of my surprise, WN directly referred me for their new grad Data Scientist position, which technically hasn’t been posted yet, so I again front ran the market (similar to what happened with Zillow).
So again, things were moving really fast: I got introduced to WN on Aug 23 Sun, and she replied on Aug 24 Mon saying she has referred me, and on Aug 25 Tue the recruiter reached out to me to hand me the data challenge. On Aug 28 Fri, I submitted the challenge, and the recruiter said they’ll follow up shortly. On Sept 2 Wed, the recruiter said I’m invited to the phone screen. On Sept 8 Tue, I had my phone screen. The next day, Sept 9 Wed, the recruiter emailed me to say I’m invited to the onsite. On Sept 10 Thu, I received my confirmation for onsite for Sept 17 Thu. The only slow down was after the onsite: it’s only after 12 days, on Sept 29 Tue, that I got the offer.
My guess was that, given that they are a small firm — a total of ~200 people, with ~15 people on the data science team, they had to be selective about offers, so they probably interviewed a bunch of people before they decided on who to give offers to. Also, it takes time to get the offer details to be approved, and the reason why other firms were faster was partly that they first gave me the offer without offer details. For the Quora case, the offer came with the details, which is probably also why it took longer.
On the content front: I’d say Quora’s interview feels quite different from Uber and Zillow’s, and closer to Google’s — it’s less business case studies, more data science. I think I pulled it off because (1) I very seriously practiced data manipulation in Python with real data sets and (2) I used their products heavily before the interview and thought hard about how I would improve it what I don’t like about it and (3) I used my on-the-fly problem solving ability to fill the gap when my knowledge was lacking.
The only firm that I both did not have a strong connection and whose process went very smoothly was Amazon, which is probably because (1) the system + recruiter were good and (2) given their long history of hiring economists, econ phds (especially from top programs) have a very easy time getting noticed.
Amazon
You would have guessed getting referrals for Amazon is the easiest because every year there are people from Stanford going into Amazon, but to my surprise, it turned out to be the hardest, precisely because there were too many connections:
As mentioned before, RE held a ‘economists in tech’ panel, and I asked her for the contacts of the 5 panelists. I contacted the economist representing Amazon, IA, and had a great chat with her in July — she answered all my questions in detail, and offered to refer me when I’m ready to apply. However, by the time I sent her my CV on Aug 22 Sat, she didn’t reply after a few days. I followed up on Aug 25 Tue, and she still didn’t reply. The hardest thing about referrals/asking for favor is that it’s not good to ask multiple people to do the same thing for you. Now that I’ve asked her for the referral, and she hasn’t replied, I honestly don’t know what to do. (She finally replied in mid Sept… turns out she just missed my email)
At the same time, I was talking with AX, a GSB phd who graduated that year who ended up in Amazon, about something completely unrelated — J1 waiver (more on this later), so I asked him for a referral for Amazon on Aug 26 Wed. He said he’ll ask the recruiter how to do it, and got back to me on Aug 28 Fri to say that I need to give him the job ID that I’d like to apply for. The problem was: there are literally thousands of jobs on Amazon’s career website under the econ category, and after narrowing things down, I found 3 job postings with basically the same content, i.e. new grad Economist. I replied him on Aug 29 Sat, explaining the situation, and he replied on Aug 30 Sun saying that he will go and ask his colleagues what to do, and again never got back… (my guess is that, Amazon work-life balance is so bad that, people are too busy)
Luckily, at the same time, I asked Anthony to introduce me to RN, former GSB PhD who now works at Amazon as an Economist (both AA and AX mentioned to me that they talked to RN) on Aug 27 Thu, and we talked on the same day. After the chat, I emailed RN my CV, and he forwarded it to the recruiter he worked with. A few days later on Sept 1 Tue, the recruiter reached out to me. — I finally got a referral in trial #3.
The rest of the Amazon experience has honestly been pretty smooth: On Sept 2 Wed, I chatted with the recruiter, who sent me the link to the job posting I should apply to. On Sept 9 Wed, the recruiter reached out to schedule my phone screen — I was a bit surprised by how long it took for my CV to pass the ‘review’ stage, which shows, again, if your CV was identified by a hiring manager to begin with (Uber and Zillow case), your process will be a lot faster. On Sept 10 Thu I got an email confirmation for my phone screen on Sept 15 Tue. After the phone screen, the next day, Sept 16 Wed, the recruiter called to say I passed, and that she will share my profile with a few hiring managers .
Again, as I said before, if you started with a hiring manager being interested in your profile, things would be faster: for Amazon, if you entered the generic new grad pipeline, then you need to go through a ‘team matching’ process between passing the phone screen and the onsite (unlike Google, which does team match after someone passes the onsite). (I’ve had a friend who re-selects the team after he passes Amazon’s onsite, but my understanding is that you need at least a hiring manager expressing interest in your profile to have the onsite)
On Sept 21 Mon, the recruiter called to say a hiring manager has expressed interest and that I’ll be invited to onsite. On Sept 22 Tue, I got an email confirming my onsite on Oct 2 Fri. (I was a bit annoyed by how late it was, as I have already expressed my time constraint with the recruiter, but it is what it is.) On Oct 5 Mon, the recruiter emailed to say there’s no offer.
On the content front, the Amazon interview is heavily behavioral — for most other firms, they had at most 1 full interview during the onsite dedicated to behavioral questions. For Amazon, it’s almost the other way around. For their ‘technical interviews’, it’s basically the set of causal inference techniques econ phds know about, but it’s very much on the practical side, i.e. if you’ve only heard of it/learnt it on the textbook, you might have issues answering some of the follow-up questions — they ask very detailed follow-up questions.
The two painful experience I’ve had in my recruiting journey were Facebook and Google, for how slow things were:
To be fair, with Facebook, I made a mistake myself: despite how much OR (previous Stanford econ phd, now at Fb CDS, one of the few people who replied to my LinkedIn cold messages) told me to apply for Novi (Facebook’s cryptocurrency team), I still only applied for the Core Data Science (CDS) team to begin with. (That was mostly because I heard that Novi was specifically looking for a monetary/finance person, which I’m not. — it later turns out I was totally wrong. Even though I wasn’t ‘monetary/finance’, I still passed the two rounds of phone screens and got to the onsite. — Another lesson I learnt: when you are on the market, don’t think too hard, just apply generously.)
After my CDS application got in, I received an email from the recruiter on Aug 26 Wed saying that ‘she will let me know if there is a fit’. After talking with RB on Aug 30 Sun, a friend of Anthony’s who was at Novi at that time, I was convinced that I should apply to Novi regardless. I added an application to Novi on Sept 2 Wed and emailed the recruiter to let her know. However, that email was never replied to.
On Sept 8 Tue, the recruiter sent me an email to invite me to complete an immigration form for CDS — that is 15 days after my CDS application got in. I told her my progress with other firms, and asked if she can help speed things up. On Sept 11 Fri, i.e. 3 days later, the recruiter emailed me saying ‘we still don’t know our headcounts for CDS for 2021 so I’m not sure we can meet your timeline’. Realizing that she might have never seen my Novi application and my email about it, I told her again that ‘I have also applied to Novi, which I know is actively looking for people.’
On Sept 14 Mon, i.e. 3 other days later, the recruiter finally emailed me to say she will be setting up a phone screen for me for Novi, and asked for my preference for programming language, and shared with me what will be covered in the phone screen. (Later it turns out that the preference for programming language was not relevant for phone screen, and the content she shared was also not what was actually covered.)
After my phone screen on Sept 22 Tue, I got to know that there’s another round of phone screen. At that time I already got an offer from Uber — I started my process with these firms at basically the same time, but somehow Uber and Zillow were a lot faster than Facebook and Google.
On Sept 25 Fri, I got to know that my second phone screen is scheduled for Oct 1 Thu (which later got rescheduled to Sept 30 Wed because the interviewer had a conflict and I directly contacted him to let me know my time constraint and asked to have it asap). Given how slow things have been so far, I highly doubt I’ll be able to have my onsite and know the result before my Uber deadline on Oct 9.
Given that, I decided it’s time to use personal connections again. On Sept 28 Mon, I messaged RB to let him know my deadline, and that the recruiter has been slow, and asked if he can let the team know my situation and potentially speed up.
On Sept 30 Wed, right after my second phone screen, the recruiter finally got back to me, this time offering to chat! — in the past it has always taken her at least 3 days to reply an email, and when I asked if she can help me speed up the answer has always been no. This time, however, during the chat, she first told me that I’ve passed my second phone screen, then collected my time for the onsite, then said they can definitely schedule the onsite before my Uber deadline, and can let me know my result very soon after the onsite. — again, it’s a lot more effective to have the team push it for you than doing it yourself.
The scheduling of the onsite was not an easy one — due to the Oct 9 Uber deadline, basically all the onsites were in the week before that. On Oct 1 Thu, the recruiter emailed me to say that the hiring manager, CN can’t really do next week, and asked if I can chat with him tomorrow, i.e. Oct 2 Fri. However, my Amazon onsite was already scheduled for Oct 2 Fri… But after letting them know, they eventually figured out a schedule, and my onsite was set for Oct 5 Mon.
The onsite wasn’t fun — for example, for the first question CN asked, I had to clarify at least 3 times before I understood what he was talking about. It was also clear that my research doesn’t really has anything to do with Novi at all. On Oct 7 Wed, the recruiter emailed me to say no offer.
On the content front: it is also very heavily case study — problems that they are actually solving for Novi. The main challenge is, they don’t have a product yet — at least at the time of my interview, so it’s not like Quora where I can use the product in advance to prep for the interview. I’d say if you have a background in finance/macro/crypto/development, it would be easier; just being very good at grasping information and solving problem on the fly also helps a lot.
For Google, the team I was most interested in to begin with was their Economist team. However, EN (Stanford econ phd who just joined Google as an Economist) told me that they were not hiring at the moment. As a result, I applied for the ‘Data Scientist, PhD University Graduate’ position on Google’s career website.
After not hearing back for a few days, I realized that I really should have asked EN for a referral, even though he’s not on the same team. Luckily, their system was set up in a way that ‘retrospective referral’ can happen — EN referred me internally on Aug 21 Fri, and it got tagged to the application I already submitted. The next Mon, Aug 24,the recruiter reached out to me to start the process.
Google’s recruiter has always been responsive, but the problem is that their recruiting system is the same as the US political system — too many checks and balances made things extremely slow. For example, they have a rigorous ‘hiring committee’ system, where a group of qualified data scientists (or engineers) review phone screen or onsite performances of data science (or engineering) applicants.
If it’s your interviewer making the decision on whether you pass or fail, you can get the decision on the day of the interview; but if it’s the hiring committee, then you get the decision the next time they meet — and they meet once every two weeks — from my phone screen on Sept 9 Wed to when I got to know the result on Sept 23 Wed it took 14 days; from my onsite on Oct 6 Tue to when I got to know that I passed the onsite on Oct 27 Tue it took 21 days!
On the content front: the phone screen was the weirdest DS interview I’ve had. I don’t even know how to describe it. It’s fun, but also hard. All that I can say is that it’s very technical, i.e. very not business case study. I think it selects for people who are book clever. However, a friend of mine who also went through Google’s DS pipeline had a completely different experience — her interview was very standard. So I honestly don’t know what’s going on with Google… The onsite for me were a lot more standardized, and very similar to the one I had with Quora — a lot of metric definition questions/how do we think about how to measure something/what data we should collect type of question.
But the struggle just began: for Google, the team matching process happens after you’ve passed the HC ‘hiring committee’ (i.e. onsite). The HC result is valid for a year, i.e. as long as you get a team match in the year following you passing your HC, you’ll get an offer. But the problem is, the team matching can take forever, e.g. 6 months.
At that time, I’m already at a relatively late stage of the negotiation with Uber and Zillow, and they won’t wait for me for much longer. I clearly can’t just sit there and pray that a team match will drop on my head.
Again, it’s time to use my personal capital — On Nov 7 Sat, I posted on WeChat asking if anyone else is also in Google’s DS pipeline this year, and if so how long did their team match take. Soon after the post, someone a few years senior to me when I was an undergrad at HKUST reached out to me to say she knows 2 people who were senior to her at HKUST who are now Data Scientists at Google, and introduced me to them. I chatted with both of them on Nov 8 Sun. On Nov 9 Mon, one of them told me that he has found a team that is hiring — Display Ads, and that the hiring manager will reach out to the recruiter to follow up. On Nov 10 Tue, the recruiter told me a hiring manager has expressed interest, and will be setting up a team matching call for me.
Before I used my personal network, I had 11 good days with no update from the recruiter. As soon as I made that post on 朋友圈, I got a match in literally 3 days.
All that I want to say with the stories is that, relationship/network/referral is really really important in job hunting, or more precisely, anything in the real world. The most common mistake PhDs make is to treat the industry market the same as the academic market, thinking that as long as they apply to 300 jobs on the AEA website they’ll end up with a good job. Yes, indeed there are firms that hire through the AEA and will interview you as long as you’re a econ phd or a econ phd from top 10, like Amazon or Uber, but if you’re interested in tech in general, and would like offers from firms other than the ones that go to the AEA, then applying by yourself and aggressively networking is really really important.
