Job Recommendations & NLP processing on Job ADs

Arbetsförmedlingen uses Hopsworks to quickly serve real-time predictions of suitable job postings for approx. 700,000 new job seekers per year in Sweden (growing by 18% from last year). They also use Hopsworks to identify job announcements that contain discriminatory text.

The Aim: 

Arbetsförmedlingen lacked a highly available production environment. They were only using open source tools, glued together and supported by the team itself. To manage and orchestrate workflows and processes around AI, they needed a platform that would allow GPU access and scheduling for both training and serving models, as well as serve online data for recommendations.

Why Hopsworks?

They used Hopsworks as a model serving platform that leverages GPUs to serve predictions on large language models.

They identified Hopsworks as a scalable platform that offers online and offline models. Plus the platform supports data lake integration with Hortonworks HDP, Cloudera CDP, relational databases and object stores. It also offers multi-tenancy and GDPR-compliant data sharing.

Results

Centralization and collaboration. The Data Scientists can work with modern libraries developing, creating feature pipelines and developing AI models in a structured manner.