#26b: MLDublin meet Microsoft @ Microsoft One

Over the last 5 years AI has come out of R&D and become a mainstream tool aimed at augmenting human intelligence. As AI applications become integral parts of business processes, and soon mission critical processes, AI models and algorithms are subject to the same sort of pressures as applications - they need to be refined and modified without any downtime, they need to be audited and logged and they should only be updated if it offers a measurable improvement over the previous variant or version. There is an urgent need for application of agile development and DevOps into data science: we need Agile AI. Andrew will discuss what can be done to make AI more agile, keep data scientists focused on what they love, and build robust service delivery approaches to AI.

Using deep learning to remove cranes from city landscape pictures Learnings and outcomes from my cooperation with a customer on building a deep learning solution to automatically remove objects (cranes) from pictures