The author shares insights from their experience interviewing for ML/AI engineering roles, emphasizing the shortcomings of traditional ML system design interviews. They conducted around 70 interviews across 15 companies, receiving seven job offers. One standout moment was a candid discussion with a CTO regarding the ineffectiveness of the ML system design interview format. The article identifies several failure modes in these interviews, such as superficial questions dressed in ML terminology that fail to probe candidates deeply, and scenarios lacking sufficient context for meaningful discussion. The author highlights the tendency for interviewers to focus on niche areas of their own experience, which can lead to a disjointed understanding of the broader system. They point out that vague questions often fail to stimulate productive dialogue, using examples of good versus bad scenarios to illustrate the point. Outdated problems, like asking candidates to design solutions without considering recent advancements in ML techniques, further detract from the interview process. Instead of focusing on rederiving established algorithms under time constraints, the author suggests assessing candidates' ability to learn and implement solutions in a reasonable timeframe. A well-designed interview loop should evaluate a candidate's practical skills and adaptability, rather than their ability to regurgitate theoretical knowledge. The piece concludes with a call for companies to rethink their ML interview processes, emphasizing the need for interviews that genuinely reflect the skills required for ML engineering roles.