Customer
Leading media research company with more than 25 years of work and research on the advertising and media monitoring market.
Challenge
Develop and implement a new DS platform in order to increase scalability, transparency of the ML models development process and reduce the time of bringing research algorithms into production use.
Project Results
Through this project the Client has received a scalable and manageable space for ML models development that enables rapid engagement of internal teams of data scientists with the ability to evaluate the results of their work. Using this platform, the company will also be able to engage external ML teams promptly and with minimal effort to increase the number of jobs being worked on and models being developed at the same time. In addition, data scientists received access to a centralized catalog of ready-made pipelines simplifying the subsequent development of models by reusing ready-made components.
Business Value
Reduced time-to-market for ML models and operational risk related to model deployment issues, simplified experiment tracking, 3x decrease of effort required for onboarding of external DS teams.
Tech Stack
Hadoop stack, Kubernetes, Kubeflow, Minio, Jupyter, Gitlab, ArgoCD, Airflow.
Leading media research company with more than 25 years of work and research on the advertising and media monitoring market.
Challenge
Develop and implement a new DS platform in order to increase scalability, transparency of the ML models development process and reduce the time of bringing research algorithms into production use.
Project Results
Through this project the Client has received a scalable and manageable space for ML models development that enables rapid engagement of internal teams of data scientists with the ability to evaluate the results of their work. Using this platform, the company will also be able to engage external ML teams promptly and with minimal effort to increase the number of jobs being worked on and models being developed at the same time. In addition, data scientists received access to a centralized catalog of ready-made pipelines simplifying the subsequent development of models by reusing ready-made components.
Business Value
Reduced time-to-market for ML models and operational risk related to model deployment issues, simplified experiment tracking, 3x decrease of effort required for onboarding of external DS teams.
Tech Stack
Hadoop stack, Kubernetes, Kubeflow, Minio, Jupyter, Gitlab, ArgoCD, Airflow.