Scaling AI across Nubank’s 100M+ customers
In this episode of Risk and Robots, Maximilian Eber, Co-Founder and CPTO of Taktile, sits down with Henrique Lopes, Head of AI at Nubank, to talk about what it really takes to put AI into production in a highly regulated environment. Henrique shares how Nubank’s first use case—accelerating code migrations—evolved into customer-facing applications, why they decided to build their own models on transaction data, and the trade-offs between rules, ML, and LLMs in banking. They also cover practical lessons on measuring success, running experiments, managing hallucinations, and knowing when to fine-tune versus train from scratch.
Risk & Robots is our ongoing podcast series where we dive into the future of risk, AI, and financial technology with industry leaders. Stay tuned for upcoming episodes as we continue exploring how innovation is reshaping the way businesses think about risk.
Follow Maximilian Eber on Linkedin for the latest tools, news, and technologies that are shaping the future of automated decision-making in finance.