Our systems run standalone and low-level diagnostics information is available, however not easily accessible to customers. In addition these require interpretation which is not present at this moment. Reliability is key, hence any intelligence which can help to detect and prevent issues is very valuable.
Use-case and problem description:
Conference systems are considered mission critical, (potential) downtime has impact on a lot of (important) people and is costly in general. Our system is now mostly used without connection to the ouside. There is some diagnostic information available, but there are two issues: 1) access to this data is not user friendly, 2) It only enables an analsysis 'post mortem'. In addition the problem analysis takes valuable time from our support or R&D teams.
With this project we want to bring diagnostics to the next level by a) exposing (more) diagnostics data visible to the end-user, b) Propose actions to be taken given system status (can even by pro-active), c) gather info and insight on utilization and status of systems worldwide.
Investigate data which can be gathered from conference system
Central unit related: cpu load, mem usage, IO access, ….
Control PC related: cpu load, mem usage, IO access, …
Unit/cable related: link losses, packet losses, temperature, voltages, …
Two step approach for the data gathering:
Expose portal for customer to upload information (like log files). Visualization and analysis tools can be developed
Check automatic upload of diagnostic data
Check data patterns, provide tools for visualization of data and patterns
Define rules/trigger to have: preventive maintenance, (early) issue detection, …
Check possibilities regarding Artificial Intelligence or Machine learning
Elastic data storage
Unsupervised learning, Anomaly Detection and Sequence modelling
Cloud platform: AWS (kinesis), Google (Pub/Sub) or Microsoft IotHub
Nature of the work
Specialty: AI / Machine Learning
Type of work: Research: 25%, Implem.:50%, Experim.: 25%