Multi-stage interview process to fintech

The multi-stage interview process to join the fintech company involved eight stages across several teams, including a live coding round, RAG (Retrieval-Augmented Generation), and agents. The company reached out to the candidate even though they had previously tried to join. The process began with a Demo Day, where the candidate received an offer in a single day after answering questions about Python concepts such as TypeVar, Protocol, mutable default arguments, and the Global Interpreter Lock (GIL).

During the second part of the interview, the candidate faced window-based tasks. The first task involved creating a rate limiter using a sliding window. Initially, the naive approach filtered the entire request list with a time complexity of O(n), but an optimized solution using a deque reduced the work per request to O(1). The candidate then discussed how machine learning CI/CD differs from regular CI/CD, highlighting the additional components in ML CI/CD, such as versioned data, models as artifacts, and quality metrics like accuracy, F1 score, and latency.