Designing Intelligence for Realtime Autonomous Systems
AI research is advancing at tremendous speeds. It seems as if every day there is a breakthrough in a certain field. While these breakthroughs can come from fundamental changes in model structures, essentially it often boils down to an increase in computational cost. A CNN with millions of parameters being compared to a linear regression for the same use case is something that for realtime autonomous systems borders on irrelevance since constraints such as power consumption, hardware cost and latency are often completely ignored. With the help of examples, insights will be given into why many AI use cases are still unlocked today although it does not necessarily have to stay this way.
Gilles is a passionate engineer at heart with an entrepreneurial mindset. He has spent his last couple of years working on various industrial problems in the domain of realtime machine learning and signal processing systems. With a Master in Electrical Engineer and a degree in Technology Management his focus revolves around machine learning designs required to consider special constraints that such systems are facing. Holistic designs targeted at the ideal symbiosis of hardware and algorithms such as stripped down neural networks on microcontrollers is what drives Gilles throughout his projects in the automotive, wireless and sensor industry as a developer and engineering lead.