Explore a 27-minute conference talk from MLCon that delves into the challenges of managing doomed machine learning projects. Learn how to navigate miscommunications, dataset issues, and unrealistic expectations when faced with seemingly insurmountable obstacles. Gain insights from Vladimir Rybakov's experience as he shares a case study on making the most out of a troubled project, discussing what truly matters when goals appear unattainable. Discover strategies for handling weaknesses in ML projects, managing expectations, and deciding whether to participate in challenging endeavors. Follow the speaker's journey through data recovery, modified metrics, and unexpected successes, ultimately learning how to turn potential failures into valuable learning experiences and maintain a positive reputation in the field.