Public defence in Communications Engineering, M.Sc.(Tech.) Edgar Ramos

The title of the thesis: Intelligence enablement and orchestration

Thesis defender: Edgar Ramos
Opponents: Prof. Kari Systä, Tampere University, Finland
Custos: Prof. Jyri Hämäläinen, Aalto University School of Electrical Engineering, Department of Information and Communications Engineering

The study investigates how intelligence can be enabled in connected devices. Also how to orchestrate the processes using such devices in an interoperable manner utilizing the connection network as a possible orchestration platform. One key enablement of intelligence is the introduction of the intelligence layer as a service layer for applications decoupling the algorithms from the application code. For orchestration, the exposure of APIs and capabilities is the base to enable the processes modeling and execution. 

The purpose of the study was to overcome the ecosystem frictions and obstacles to enable intelligent models in devices taking into account data preparation, training, execution, and model updating. In similar way, the enabling interworking of several devices takes into account the heterogeneity of IoT and the variety of use cases for each device. By looking at the possibility of establishing an orchestration process that doesn't require changes in the devices but instead is supported by platform services. 

The use of AI and the commoditization of intelligence have made these results highly relevant for the current systems and device application production. Also for the future possibilities of standardization of formats for orchestration of different devices for their use cases. The frameworks and tools that the research brings can be incorporated into the future design of applications, platforms, and devices, including operating systems. 

The results are very relevant for the design of the new "AI-as-a-Service" platforms and the introduction of updatable intelligence in devices. Also, the wish is to support the switching of the training data ownership from the application provider to the user. This would make it possible to export the data to be used by a different model instead of being just locked in the application itself.

Keywords: Intelligence enablement, Intelligence orchestration, AI Lifecycle Management

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53