Join itag AI forum for our October meet up which will cover two topics ‘AI-powered AD for CI/CD pipelines’ and ‘Time-series anomaly detection insights’.
Time-series anomaly detection insights
Stefano Mauceri, Machine Learning Engineer, Huawei Ireland Research Centre
After working on several anomaly detection problems I have learned some valuable lessons that I would like to share with you. I will keep the presentation simple and short such that we can spend more time discussing together. My goal is to let you take best out of your next anomaly detection project.
AI-powered AD for CI/CD pipelines
In a fast-paced software development life cycle which heavily relies on high availability and health of a CI/CD pipeline, being able to quickly detect, analyze, and resolve anomalies ensures consistent and efficient development activities and delivery of software artifacts.
RUAD (Recurrent Unsupervised Anomaly Detection) is a novel anomaly detection approach developed by our Chief AI engineer, Martin Molan. We will share the practical experiences of running RUAD anomaly detection system in production for over a year and discuss the possibilities to customize and integrate RUAD into any CI/CD pipeline to help DevOps teams to easily detect and resolve issues.
It can handle complex system monitoring data and it is the first anomaly detection approach capable of completely unsupervised training. It was initially created to monitor the Marconi 100 supercomputer and then adopted to support ITOps at Comtrade Gaming.
Key points about the solution:
– Anomaly detection has to be trained for each monitored system.
– Does not require any external tools (OpenAI, large language models, generative models) nor pretrained models.
– The log understanding is based on domain knowledge, not on large language models.
– The algorithm is trained and deployed on premise.
Domen Ferbar from Comtrade Gaming & Vera Barić from Comtrade 360 – AI-powered Anomaly Detection for CI/CD Pipelines
Domen Ferbar leads IT operations at Comtrade Gaming, an independent software provider that delivers open gaming platforms and professional services to both online and land based gaming sectors.
He currently implements and operates customer instances and dev/qa environments in the cloud and on prem.
Prior to this role, he transitioned from being a trained science teacher to managing IT for a major financial establishment. Additionally, he has experienced the exciting world of building and supporting infrastructure for a Central and Eastern Europe omnichannel retailer.
Vera Barić, Machine Learning Engineer, Comtrade 360
As a Machine Learning engineer at Comtrade 360, Vera Barić works with colleagues to create and implement real-world solutions around her area of expertise; Anomaly Detection and Casuality Analysis.
She is an avid tech enthusiast, who is constantly exploring the latest trends and engaging in continuous learning. Beyond data and technology, she is passionate about outdoor activities like hiking, rock climbing, and running.
Vera has a Bachelor’s degree in Engineering, focused on data-driven solutions and is currently pursuing an MSc in Data Science at ITS.
Alan Duggan, DevOps Architect, Hewlett Packard Enterprise
Alan leads the HPE Support Centre DevOps Platform team, who provide the infrastructure, tooling and automation frameworks, for multiple scrum teams to develop with.
He is passionate about learning new ways to streamline software delivery, improve CI/CD processes and Agile principles.
As a way of sharing the learnings from this, Alan is also involved in the DevOps Galway Meetups and DevOpsDays Galway.