Finding the right job can be a challenge, made more difficult by undefined job titles and countless job boards. In the consulting process, von Rundstedt matches these vacancies with the career goals previously developed for its clients, using specially designed algorithms. The result: job suggestions that truly fit and invite professional change. Smart Vacancy Matching at von Rundstedt was not only a Nominee for the German HR Excellence Awards but also advanced to the final round within the category of Analytics & Technology (AI in HR).
Optimize an AI Powered Recommendation Engine
A Collaboration between Intel, scieneers, and von Rundstedt
Intel Corporation is a world leader in computing innovation that designs and builds the essential technologies that power the cloud and smart devices. scieneers is an Intel Partner Alliance Gold Member, providing customized solutions and services in Data Analytics and Artificial Intelligence for various industries. von Rundstedt is a leading career and outplacement consulting firm that helps people and organizations navigate the changing world of work.
The three companies collaborated to optimize Von Rundstedt’s AI powered recommendation engine solution, which helps job seekers find the best career opportunities based on their skills, interests, and goals. The solution uses natural language processing (NLP) and machine learning (ML) techniques to analyze job descriptions, resumes, and feedback from users and employers. It then generates personalized recommendations for job seekers, as well as insights for employers on how to attract and retain talent.
Currently, to feed the recommendation engine, 1.2 Million embeddings of job descriptions, title and meta data are created on a weekly basis utilizing a customized BERT-model. The initial situation covers this compute intense workload by means of a machine learning pipeline deployed on an Azure Virtual Machine which is powered by Intel Xeon E5-2690 v3 (Haswell) CPUs. The embedding of 1.2 Million samples takes around 15 hours to complete and is thus quite expensive as well as time consuming. Therefore, an optimization is highly welcome and of great value for von Rundstedt and their users.
The optimization process involved using the Intel® AI Analytics Toolkit, which is a comprehensive suite of tools and libraries that enable developers to build, analyze, optimize, and scale data-centric applications across diverse architectures. The toolkit includes Intel Distribution of Modin*, which accelerates data processing by distributing workloads across multiple cores or nodes; Intel Distribution for Python*, which delivers optimized performance for Python applications; Intel Data Analytics Acceleration Library (Intel DAAL), which provides highly optimized algorithms for data analysis; and Intel Optimization for TensorFlow* as well as Pytorch*, which enhances the performance of TensorFlow/Pytorch workloads on Intel hardware.
By using the Intel Intel® AI Analytics Toolkit, the three companies were able to achieve a significant improvement in performance for the AI powered recommendation engine solution. The solution achieved up to 9.3X speedup when running on 4th Generation Intel® Xeon® Scalable Processors compared to the baseline configuration. This means that the solution can process more data faster, generate more accurate recommendations, and serve more users with lower latency and cost.
Taking the results into account, a migration from the current to the optimized solution with a deployment on 4th Generation Intel® Xeon® Scalable Processors, which are designed to deliver breakthrough performance for data-centric workloads. The 4th Generation Intel® Xeon® Scalable Processors feature Intel Deep Learning Boost (Intel DL Boost), which accelerates AI inference performance; Intel Advanced Vector Extensions 512 (Intel AVX-512), which boosts performance for compute-intensive applications; and Intel Optane Persistent Memory 200 Series, which provides large-capacity and persistent memory for data-intensive workloads.
The collaboration between Intel Corporation, Scieneers, and Von Rundstedt demonstrates how AI can create value for businesses and customers across different domains. It also showcases how the Intel® AI Analytics Toolkit and the 4th Generation Intel® Xeon® Scalable Processors can enable developers to optimize their AI solutions for performance, scalability, and efficiency.
Artificial intelligence (AI) is transforming the way businesses operate and deliver value to their customers. One of the key applications of AI are recommendation engines, which provide personalized suggestions to users based on their preferences, behavior, and context. Recommendation engines can enhance customer satisfaction, loyalty, and retention, as well as increase revenue and profitability.
However, developing and deploying a high-performance recommendation engine is not a trivial task. It requires a lot of data, computing power, and expertise. That’s why Intel Corporation, Scieneers, and Von Rundstedt joined forces to optimize an AI powered recommendation engine solution that leverages the Intel® AI Analytics Toolkit and 4th Generation Intel® Xeon® Scalable Processors
As a service provider and developer of data-driven applications, we look forward to continuing to increase the added value for our customers in cooperation with Intel®. Especially with regard to the machine learning operations of large LLMs, we look forward to future technological developments and are pleased to be able to share further experiences in dealing with state-of-the-art technology. At this point we would particularly like to thank all partners involved in the cooperation, which led to the results shown here