10 years of engineering at Nubank: Lessons from scaling to 122+ million customers (Category Architecture)
Recently I read it. article The evolution of NuBank 10 Lucas Cavalcant, Distinguished Software Engineer and Senior Architect This story of evolution seemed interesting to me, so I decided to talk about lessons learned.
Leadership beyond code Technical leadership requires not only writing great code, but also the ability to communicate ideas clearly, inspire a team and guide them to a solution – especially when you don’t have the formal levers of a leader. (This is a feature of staff + IC branch). As the role of an engineer grows, he has to engage in strategic planning, coordinate several teams and make major technical decisions. We also have a branch in T that I'm talking about a lot. told.
Career path (technique vs. management) In the early years of Nubank, engineers worked in survival mode - problems arose daily, and everyone rushed to fix them. Only to 2017 The first managers were introduced, and many (including Lucas Cavalcanti) Then they realized that they were closer to the track IC. (individual contributor) And the company took that into account. 2018Lucas became the first Principal Engineer, and recently rose to Distinguished Engineer.
The evolution of architecture under hypergrowth Scaling from scratch (into 2013 year) Tens and hundreds of millions of customers required constant architectural changes. Originally, Nubank was written in Clojure with Datomic and deployed monolithically in AWS. (CloudFormation + AMI). Details in speech 2017 years at QCon. And the use of AWS here As the load grew, they switched to Docker and Kubernetes, sawed the monolith into microservices, and later created a core banking platform capable of supporting many products in different countries. Details in article 2019 microservices year International expansion brings new demands (Local regulations, languages, etc.) But not all large-scale solutions have been durable: the data sharding introduced in 2016At first, it helped, but in the end it got into the physical limits. (AWS did not have time to issue new servers). Today, as the speaker notes, fundamentally new architectural approaches are needed for further growth.
Standards vs Innovation At the dawn of Nubank, engineers themselves banded together in a "guild" to experiment with technology and tools beyond basic product work. This approach created a culture of innovation, but it did not scale well. The company now has separate platform teams building common tools and infrastructure, allowing product teams to focus on their features. Platform teams need to invent new approaches, tools, and operational practices, as Nubank often lacks ready-made solutions.
Team performance without burnout Speaker divides his path into two stages: early start-up years at a frantic pace (permanent "fire regime") A more structured present, where processes are built that do not allow endless extinguishing of incidents. His advice: at the start of a career or product, you should give your best, but it is impossible to constantly work this way. Over time, Nubank has consciously improved the work-life balance of teams.
Looking to the future (AI and fintech) Artificial intelligence technologies are named as the next big breakthrough, primarily large language models. (LLM). They can dramatically increase developer productivity, but there is a risk of generating more change than teams can meaningfully consume and follow. The challenge is to find the right cases where AI really provides stable value, not tons of extra code. At the same time, breakthroughs are taking place in fintech itself: for example, Pix, an instant payment system launched in Brazil, has radically changed the way people handle money.
#Software #Engineering #Management #Architecture #Processes #Staff #SRE #DevOps