ICHEC completes project with Opening, MS and Nvidia on talent discovery
Opening.io, an Artificial Intelligence (AI) startup based in Dublin, has have recently completed work with the Irish Centre for High End Computing (ICHEC) in partnership with Microsoft Ireland and Nvidia on a proof of concept project aimed at increasing software performance by porting CPU based software to GPUs.
Opening.io operates in the talent and recruitment space, incorporating linguistics algorithms to identify patterns within the structure and phrasing of job posts and CVs, converting them into data points to match candidates to suitable jobs. These business critical processes involve aspects of machine intelligence deploying deep learning networks.
ICHEC, in partnership with Microsoft Ireland and Nvidia, tested diverse workloads allowing for insights in terms of cost/performance gains from GPU nodes in different scenarios, ultimately resulting in increased performance, representing a faster completion time and a lower operating cost for the company.
According to ICHEC’s computational scientist Dr. Paddy Ó Conbhuí, who worked on the project, “one big advantage of using GPUs over CPUs is that they can provide very high ‘bang per buck’ in terms of performance and energy usage. This is a real issue as supercomputing heads towards the exascale regime, where current projections suggest incredible levels of power will be required.”
“However, GPUs can be trickier to write efficient applications for, and difficult to predict the performance which can be achieved,” he added. “As a result, it is often necessary to write a proof-of-concept GPU port to determine this, so to direct a business decision.”
Essentially, Opening.io successfully deployed its next generation inference engine enabling much faster (by a factor between 5-10x) intelligence and functionality delivered in real-time and at Fortune500 scale: neural parsing of resumes, talent discovery, data augmentation and enrichment, intelligence and forecasts. By fine-tuning interactions between software and GPU’s computation power Opening is now able to handle vastly increased load while using the same underlying hardware to power the platform.
In recent years the market has experienced a substantial increase in the uptake of GPUs - a welcomed transition into High Performance Computing (HPC). Since the early days of application in computer graphics the benefits of these processors have outgrown expectation, from a general graphics accelerator to the go-to tool for performance optimisation, servicing new domains and sectors from financial and banking to engineering and real-time video processing.
ICHEC’s senior data scientist Dr. Bruno Voisin, who led the partnership, stated that “in a production context, increased computing performances are only worthy if they provide a benefit to the business. While GPUs allow faster processing of more data, such cloud resources are more expensive to rent than standard CPU nodes, hence the need for an analysis of how the performance gains and cloud computing costs scale in different scenarios.”
“These results can then guide the business decisions in terms of dynamic cloud resources provisioning,” he said.
Ultimately, proof of concept projects like these are a great opportunity for SMEs to assess the technology while demonstrating business benefits through the increased performance of their solution running on GPUs. Moreover, such projects demonstrate to SMEs innovating and working on data-intensive AI and machine learning processes the need to explore the potential benefits gained from High Performance Computing.