Goodbye & much ❤ Spotify
After four cherishable years at Spotify, in what was a very tough decision, I decided to leave Spotify last week. Spotify is one of the most caring employers around, and I am filled with gratitude for the rich experience it has given me over the years.
After spending years researching search and information retrieval during my PhD, Spotify gave me a chance to pivot my research a bit and focus on recommendations. It was during my time at Spotify that I realized the attractiveness of multi-stakeholder platforms, the emergence of creator economy and the role responsible algorithmic decisioning plays in developing sustainable platform ecosystems. In my new job, I am excited to continue to focus on these topics but more on this in another post.
At Spotify I felt empowered to conduct high quality research, impact product strategy, develop and deploy state-of-the-art ML models, own production systems, contribute to many production launches touching 400 million users, and have fun while doing all this. Spotify has been a very special place: I was fortunate to be a Spotifier during special times: witnessing our IPO in early 2018 💹, growing our user base from 150M MAU to over 400M MAU 🚀🚀, launching personalized playlists loved by millions, becoming the top podcast destination in the world🎙, and introducing industry changing creator marketing tools. So much has happened in so little time — all thanks to the tremendous talent and leadership we have here.
While I have a tonne of learnings and experiences to share, here’s a short list of five things I especially liked that are worth highlighting (in no particular order).
- Rich set of problems. Extremely rich. I’m still amazed by the variety of problems one gets to tackle at Spotify. If one looks at some research tracks here, or some specific topics mentioned here, here, or here — it is no overstatement that at Spotify there will be some team or the other who would be facing these problems, and would be interested in collaborating. Just in four years, I have had the chance to work on a wide array of topics:
→ user understanding for personalization (e.g. user intents, rewards, user receptivity)
→ evaluation & measurement (e.g. counterfactual estimators, causal impact, interventions)
→ search (e.g. query understanding)
→ multi-objective decisioning in marketplace (balancing, consumption shifting)
→ ML modeling & recommendations (bandits, multi-task representations)
→ recommendation aspects (e.g. diversity, cross-market)
There is no dearth of interesting and challenging problems to work on, and most of these problems are not just Spotify specific, but are very relevant for the wider tech industry. Getting an exposure to such rich problems is one of the best things that can happen to an industrial researcher’s career. - Growth & fluidity of roles. I joined as a researcher, working on one topic with one goal with one squad. In the next 4 years, I grew to lead tech across multiple squads with people from multiple roles, and multiple missions, going after multiple objectives. The flexibility in the scope of responsibilities and the circle of impact was a key hallmark of my time at Spotify. I could dive down the rabbit hole as a data scientist digging up findings and insights, use them to inform modelling choices as a researcher and implement a first version as an MLE. Immediately after, I could start roadmapping with our amazing PMs on what a testing and deployment roadmap might look like. And in a few weeks time, I would then work with partner teams that rely on our core engineering system to ensure they’re able to use the system and get their metric wins from it. This multi-hat experience has been extremely enriching, and enlightening.
For an individual, there is tremendous room to grow and have wide impact. While opportunities won’t be served to you on a platter, but they are there if one looks for them. - Industrial research. How do you push the state-of-the-art via research while and at the same time mature your production systems to needed levels of sophistication, all at the same time? It’s often hard for any research org to maintain the right balance between research and product impact. In our attempt to get this right, we managed to maintain a good portfolio of projects: both short term impact focused, and long term strategic. This allowed us to not only maintain extremely close collaborations with product teams, but also ensure alignment and influence with senior leadership . As a researcher, the social capital I managed to earn via close partnerships with product has been a big hallmark of my time at Spotify. (how to do this right, and some key lessons learned is a topic for a longer post for another time)
- Contributions to the community. I feel proud to have been part of the team that has published papers, released datasets, written blog posts, organized TREC tracks, released research code, hosted academic visitors, organized workshops and given tutorials — all of this within a short span of 3 years. I personally believe it is important for companies to help maintain a strong bridge between academic research and industry; and we spent a lot of efforts in making high effort contributions from our side. To list a few:
→ we openly published (e.g. Spotify publications & research blog posts)
→ we released pretty big large scale public datasets (e.g. million playlists, million sessions, 100k podcasts)
→ we rallied academic research on podcasts via TREC Podcast track
→ we gave tutorials at top tier conferences (e.g. marketplace recommendations at KDD, user engagement at WSDM, grounding metrics for human ML systems at NeurIPS) - Inclusive culture. I cannot talk about my time at Spotify without mentioning the extraordinary work culture I got to witness. Spotify maintains a good balance between employee autonomy and accountability. Still heavily influenced by the Swedish culture (we still call out get togethers fikas), it offers a high trust environment. A fun work environment, and happy colleagues foster very positive vibes.
While the list of people to thank is too huge, and consists of amazing folks from Tech Research, Mixer, Creator Studio, MIQ, Home, Search and many other teams & product areas, it’s worth calling out few specific people: Fernando Diaz, who hired me back in 2017 into this amazing company, Mounia Lalmas, who has been my manager for most of my time here, and a constant guide and mentor for me and many many others, and Jeremy Ball & Chris Harris for being exceptional engineering leaders, guiding and shaping all my tech advancements.
If you’re interested in solving interesting, challenging problems in a fun environment, I’d highly recommend you look at open positions, and consider joining the band. Happy to put you in touch with relevant Spotifiers.
As a user, and as an ex-band member, I ❤ Spotify. I continue to be the biggest advocate for Spotify, its people, its leadership and its mission.
