Price: 0
Number of applications: 3
09.01.26 (inclusive)
In agreement with the Contractor
Finished product
ICT tasks
Robotics
Information processing and transformation
Software/ IS
The current system lacks a mechanism for early and automatic detection of failures and degrades in the ML ranking algorithm, which is why problems are identified after the fact by business metrics, require manual diagnosis, and lead to financial and product losses.
1. Alerts are triggered correctly in case of synthetic failures 2. No false alerts during normal operation 3. The ability to quickly locate the source of the problem 4. Convenience of interpreting alerts
Kim Sergey Romanovich
Purpose and description of task (project)
The marketplace uses an ML ranking algorithm that forms the order in which products/services are displayed in search and listings. The algorithm directly affects the user experience and key business metrics (CTR, CR, GMV). The purpose of the task is to develop a monitoring and alerting system that will allow: - promptly detect the degradation of the algorithm, - automatically signal failures, - help in determining the possible cause of the problem (data, model, infrastructure). The alert system should cover: - model input data - the behavior of the model in production - online business metrics - the technical condition of the ranking service.
Whatsapp
