My tour of duties at Coinbase
Last week marked the end of my tour of duties at Coinbase where I built many systems and teams from ground up (data, risk, tools, identity). It was an exhilarating three and half years and Coinbase is quite a rocket-ship that I am immensely grateful to have had a front row seat on.
The core values at Coinbase (efficient execution, clear communication, positive energy & continuous learning) are the crucial pillars that have made the company successful. As I reflect upon the last 3.5 years, I thought I should highlight these values as I experienced them throughout my journey at Coinbase.
Positive energy: I am grateful for the opportunity to do the best work of my career at Coinbase. I got to build some cutting edge technology and amazing teams (data, risk, tools, identity) during my time here. However, what I am most thankful for is that Coinbase taught me how to survive and win. We had many critical moments during my tenure that could have broken our trust with regulatory partners or users but we survived and won over all of those issues and built an incredible business due to only one thing: the positive energy and amazing resilience of the people here. Here are a few examples:
- At Coinbase, we have a pretty hard payment fraud problem due to the instant and irreversible nature of cryptocurrencies. Over the last several years, we’ve built a sophisticated system based on machine learning that detects how likely it is that someone is using a stolen payment instrument to purchase cryptocurrencies. We then use a user’s risk score to determine their purchase limits or withdrawal restrictions such that we can balance fraud loss to Coinbase vs. undue friction to good users.
- When phone porting attacks were affecting our customers, we built a risk model that detects potential account takeovers and delays logins or withdrawal of funds until the user can do a step-up authentication.
- During the crypto bull run of 2017, we had huge backlogs in Identity document verification. We had to turn away several customers. We quickly overhauled our verification infrastructure and went from a fully manual process to one where an automated system does a first pass at annotating and validating IDs followed by a fallback to the manual process.
Efficient execution: My favorite moments at Coinbase are from when we were resource constrained during the crypto winter of 2015-2016. We embraced our core value of efficient execution then by doing things very creatively. It was hard to recruit top talent unless you showed them what it was like to truly work here. We would invite our candidates over for a few days to a week to do a paid work trial — they would work on a problem that’s similar to the job they would do when hired. This worked well for both us and the candidates: candidates would ship a project during their work trial and get to see it’s impact first-hand. Many great features and projects had humble beginnings from a work trial. For instance, the very first machine learning pipeline and model at Coinbase had its genesis in my work trial. Given the amount of time and commitment required from candidates, we’ve since then moved away from work trials to a day long on-site interview. These on-site interviews still maintain the same ethos as our work trial — for example, data candidates are given access to real production data to build machine learning models or gather insights from.
Clear communication: Coinbase values continuous feedback and most importantly being kind over being right. We look for this value right from the interview stage. For instance, after a machine learning engineering candidate builds a model, they walk the entire machine learning team through their code and the results. The focus here is not to see whether engineers are public speakers but to find out: 1) if they can explain their thought process clearly and 2) whether they are receptive to feedback and 3) thereby can work together with their team. Similarly, our data analytics interview process involves a data insights exercise that the candidates then walk us through over a 15 minutes presentation. We honed and refined this presentation aspect over time e.g. to make it more natural, instead of requiring candidates to present via a projector in a conference room, we would often huddle over their laptop. This interview process has served us well as we think being able to succinctly explain insights to your audience is critical to a successful career in data science.
Continuous learning: Brian and the leadership team think of Coinbase as a crypto university, where you come to learn about cryptocurrencies and when you graduate from Coinbase, you spread the knowledge to the world to bring a meaningful change. That’s why you see so many Coinbase alumni who have gone on to building companies and funds in the crypto space. Personally, I learnt a ton about cryptocurrencies, security and payments while at Coinbase. I was also surrounded by incredible learners and one of my favorite memories of Coinbase is being able to light an initial spark to many of these learners and then witnessing a nuclear growth. For instance, Justin Mart was a theology major and he first picked up SQL and later machine learning to become our first data scientist. Tom Boice was an English major and was managing our anti-fraud operations team. He picked up coding in his spare time and became one of the first anti-fraud and later identity engineers at Coinbase. Jesse Posner was a lawyer turned coder and was the first internal tools engineer on the team.
Next Steps: As I went through this amazing ride at Coinbase, I realized that I enjoy building things from scratch. I am thankful to all my mentors at Coinbase — Brian Armstrong, Jeremy Henrickson, Dan Romero and Balaji Srinivasan — who gave me this incredible opportunity. The teams I built are executing well with new leaders in place. If you or someone you know is thinking of a career in cryptocurrencies, then do check out Coinbase careers page. To the moon!
Stay tuned for my next play.
