Roman Kastusik

Adoption of AI in IoT applications currently presents multiple challenges. Everything should start by identifying the most important problems and opportunities for end users. In high-performance sailing, a clear opportunity was identified: improving the autopilot. This talk explores the technical efforts and achievements using recurrent neural networks, deep reinforcement learning, and optimisation techniques, combined with domain knowledge to produce an autopilot that can operate either as a digital twin of the sailor, or develop its own style and strategy. Discover how solution co-creation between end user, subject matter experts, data scientist, academics, among others, takes place using an agile approach. You will see how implementation of solution, which spanned culture and process changes (not just technology) was undertaken (digital transformation program).