Perhaps I'm old and tired, but I always think that the chances of finding out what really is going on are so absurdly remote that the only thing to do is to say, “Hang the sense of it” it and just keep yourself occupied.
— Douglas Adams, The Hitchhiker's Guide to the Galaxy
I have spent over a decade working in finance and technology.
Some History
I began my career as a statistician building models for marketing and insurance back in 2012. My foray into consumer and commercial banking began in 2014 where my focus was still on building risk and marketing models. In 2017, I left to work at Goldman as an early employee at Marcus. I was a part of the Credit team but most of my work was all in risk engineering, data engineering, and deploying statistical machine learning models1. Then I went on to build my own consumer fintech startup where I was building a personal financial management tool.
After my startup, I joined Fast where I was hired to build risk machine learning—I ended up building out their data and machine learning infrastructure, building their recommender system, and leading two teams.
After Fast imploded, I joined Affirm as part of an acquihire. At Affirm, my team and I also built data and machine learning infrastructure—this time at a scale that was the highest I’d ever supported.
And so my career went from building models, to deploying models, to building infrastructure to deploy models…all for the glory of Fintech.
One might misinterpret that my career choices were more deliberate than I intended. At each step I tried to make choices that were rational but I mostly stumbled into machine learning infrastructure because I thought it was fun.
And now my time has come to join the artificial intelligence race.
I spent the past two years at Affirm managing a team that built out their feature store infrastructure (among other things). A feature store plays a rather critical role in the machine learning workflow and over the past year I became a maintainer for one of the most popular open source frameworks: Feast.
Open Source
I love open source software—it is the accumulation of knowledge, pain and suffering, and thoughtfulness forged by the hands of a few who were willing to build something bigger than themselves. That may sound like hyperbole but it’s not.
To quote Soumith Chintala, the lead author of PyTorch and AI Infrastructure leader at Meta:
PyTorch currently powers most of the world’s AI research and product – with hundreds of companies, research labs and individuals using and maintaining it. It has significant and tangible real-world impact from self-driving cars to drug discovery to cancer research to NASA’s Mars Rover to several consumer products, and that amount of real-world usage often intimidates me.
PyTorch began as a niche project using the programming language Lua and over the last ten years it exploded into one of the most powerful open source projects powering AI. Getting to see its glorious rise has been inspiring.
At a moment when LLMs, foundation models, and vector databases are dominating the attention of most of the media, I am choosing to work on the overlooked (perhaps underinvested) portion of AI: data2.
In most business applications outside of Computer Vision and Natural Language Processing, plumbing structured data is one of the biggest challenges in getting models into production (i.e., providing value to customers in live product experiences). In fact, I wrote an entire article about the challenges of production machine learning.
I believe Feast is a solution that enables companies to deliver more products powered by machine learning to their customers easier and faster. I also believe it can have the same community and impact that PyTorch does.
So I have decided to leave Affirm to focus on building more open source AI infrastructure including but not limited to Feast.
Farewell Fintech
I am sad to leave my team, a company I love, and a mission I care deeply3 about…but I believe we are living through an extraordinary moment in time. I want to be able to look back at the next decade and say I played a role, no matter how small, in shaping the future.
I believe Affirm is the ultimate Fintech and I know it will continue to scale into a colossal powerhouse—Affirm’s ability to underwrite risk, create exceptional products, and distribute them to their customers in a capital efficient way gives them massive competitive advantages compared to other Fintechs, so I am long the stock.
Note: this is not investment advice please do not buy stocks because of stuff a stranger wrote on the internet.
As I leave Affirm, I am so grateful for all of the incredible people I worked with—especially for my team4. The engineering talent is extraordinary and the impact is high.
A thing isn’t beautiful because it lasts.
I was surprised at how emotional I was to leave and that speaks to how wonderful my time was there and while I feel sad to go, I am so excited about my next adventure.
Hello Red Hat
I am joining Red Hat OpenShift AI. Red Hat is an industry leader in open source software. I wrote before about the need for AI to be decentralized (i.e., not concentrated to a single few large entities) and by working on open source and OpenShift AI, I will be able to enable other companies, startups, and engineers to build more AI.
The caliber of engineers at Red Hat is legendary and their history in empowering and amplifying open source is inspiring. I am so excited to get to work with and learn from such talented people and accelerate production AI.
Closing Thoughts
If you’re in Fintech, building in AI, working on an early stage startup, or all three please reach out—I would love to help5. I want you to be successful and I want the world to have more great products, there are enough bad ones.
The future will be powered by more software, data, machine learning, and artificial intelligence, not less. So my goal is simple: accelerate AI.
It’s time to build.
-Francisco
Some Content Recommendations
Jason Mikula shared the latest on Crypto Fraud and Money Laundering.
Nik Milanović announced Fintech Fund II and if you are a Fintech startup looking to raise you should reach out to him.
Nik also shared some incredible wisdom about unnatural product experiences.
Simon Taylor shared his thoughts about The Rise and Fall of Fast. 🥲
Alex Johnson wrote a wonderful piece about The Risky End State for Consumer Lending. I respectfully disagree with his conclusion though. I think a lot of lending and risk management can be replaced with machine learning and software (i.e., AI) but certainly there’s an irreducible human component defining which objectives to optimize. As I wrote before, lending is a math puzzle (mostly).
Post Script
The views expressed here are those of the author and the author alone, they do not reflect the views of his employer or previous employers.
Did you like this post? Do you have any feedback? Do you have some topics you’d like me to write about? Do you have any ideas how I could make this better? I’d love your feedback!
Feel free to respond to this email or reach out to me on Twitter! 🤠
For better or worse, the reputation I had was as a problem solver so I often found myself working on probably too many things. This is true in general.
I should note that I actually will work on LLMs, foundation models, and vector databases, too, but my point is that data isn’t getting as much attention as it deserves probably because it’s boring.
I have spent a meaningful share of my life trying to give consumers better financial products using technology and that’s been deeply fulfilling for me.
If any of you reading this are from my team or worked for me in the past, I hope you know that how honored I felt to have been given the responsibility and privilege to support, empower, and guide you—I never took it lightly.
Especially if you want to use open source. 🤠
end of an era, cheers to the next one