MLOps Marathon 2023 consists of three phases: Warm Up, Speed Up, and Finish Up, each with two stages. While all phases involve ML problems, they will vary in difficulty levels and scoring weights.
In stage 1 of each phase, teams will be assigned several ML problems and must train models to solve them. Following this, teams will deploy APIs on designated remote servers to serve the ML models.
In stage 2, the organizer will send batch requests to teams' APIs, and the APIs must return predictions for the data included in the batch requests. Results will be evaluated using both ML metrics, such as AUC and F1, as well as system metrics, such as response time.
Each phase will have different levels of challenge in the data sent in batch requests. This may include bad data such as outliers, unknown formats, or missing values, as well as drifted data, such as data or concept drift. Teams will be expected to continuously improve their models based on the evaluation results.
There are many resources available for learning MLOps, including online courses, tutorials, and books. Some popular online platforms that offer MLOps courses include Coursera, Udacity, and edX. Additionally, the necessary skills for MLOps Marathon 2023 have been distilled and designed by the MLOpsVN team into a course called MLOps Crash Course. Since 12 Dec 2022, the course has been offered free of charge to the community. You can register the course here with Access Code "MLOPS-MARA".
Open Factor Foundation is the community organization behind AIHUB.ML and MLOpsVN. Over the past years, AIHUB.ML has organized many AI-related competitions, generating great interest and impact in the tech and business communities such as AICOVIDVN 115M, MC-OCR 2021, and VLSP 2021, and now it’s MLOps Marathon.
Through these competitions, we commit to enable Vietnamese youth to bring inclusive and effective AI solutions to industries and sectors. We empower the youth with systematic and practical best practices from global AI industry tailored for Vietnam market.