DSC SCHEDULE

90% of ML products never make it to production. This is true, largely because the expertise to deploy ML solutions, maintain them and constantly scale them is scarce. As organisations learn about the transformational power of data&AI to product success, executive leaders and technology teams are actively seeking ways to incorporate AI and machine learning (ML) into their products and research. What technology to choose? What will the ML system architecture for production look like? What platforms to drive the ML system architecture? How can the model be deployed, maintained, monitored and scaled? Dynamic ML pipelines play a crucial role in adapting to evolving data landscapes and business requirements. In this session, we will explore the transformative power of AI and machine learning (ML) in driving business success. The spotlight will delve into practical strategies for – Designing end-to-end ML commercial scale product solutions. – Highlighting the critical role of data strategy in ML product development and guide attendees on how to develop an effective data strategy. – Discussion on how the right technology choices are made; ensuring a seamless integration of AI/ML into business operations. – Robust data preprocessing and feature engineering techniques – Designing system architectures that drives ML products to scale – Deployment techniques and MLOps setup/tuning at enterprise level – Detecting model drift and implementing retraining strategies – Managing and versioning ML models in a production environment – Real-world examples and case studies illustrating successful ML productionization – Guidance on implementing MLOps practices for businesses of various sizes and industries. At the end of this session participants will know how ML models are productionized at industrial/commercial level, monitor it, tune and perform MLOps for businesses.

In today’s world, more and more industries increasingly rely on ML to improve their business, and a large number of them require near-instant processing in a streaming scenario – from autonomous vehicles and healthcare monitoring, all the way to fraud detection and recommendation systems. How can we create systems that work with extremely low latencies? How can we make these systems scalable? How to ensure high availability without data loss? This talk attempts to answer these questions and more.

This talk will explain how we deploy computer vision models at Veriff by leveraging a mix of classic software engineering techniques combined with last AWS tooling. Thanks to this, we have cut down our model deployment time from 10 days to 1 day and inference costs by 75%