Machine learning training compute resources has been doubling every 6 months for the past 10 years, still to this day the lack of data is one of the main reasons why ML projects fail. This problem is especially significant for the “offline businesses”, where ML algorithms need smart sensors to gain access to the real world.
Significant progress has been made by the tinyML community towards both hardware and software optimisations to make evaluation and training of the ML models directly on sensor both fast and energy efficient. Wireless data transmission continues to be energy-intensive and expensive.
Together with the world-leading infrastructure companies Everynet is deploying nation-wide telecommunication networks dedicated to the low-cost and low-power connectivity for battery powered sensors. Network coverage is currently available in the USA, UK, Brazil, Indonesia, Italy, Spain, Ireland, etc.
Authors would like to present some of the implemented use cases and explore device-to-cloud energy-aware time series replication best practices in detail.
Schedule
Timezone: PST
Low-cost energy-aware sensor data acquisition at scale
Vitaly KLEBAN, Co-founder and CTO
Everynet
Vitaly KLEBAN, Co-founder and CTO
Everynet
Vitaly Kleban is a serial inventor and technology architect, active in deep tech: advanced sensor and materials technology, low-power and low-cost data acquisition, machine learning on edge, distributed systems and telecommunications. He is the co-founder and CTO at Everynet, which is the leading low power wide area pioneer and network operator, providing innovative solutions that help companies around the world solve their biggest IoT challenges and scale quickly.
Schedule subject to change without notice.