Post Doc

**About the role**: The installation of introspection or tracing tools, generally very useful for analysis, is not always compatible with production constraints. Ongoing work on event detection and correlation has led to a probe named "DynamicPerf". This non-invasive tool should continuously analyze the available data (metrics, logs) to perform a diagnosis and automatically propose instructions/recommendations. The constitution of a knowledge base of signatures of known problems should make it possible to detect the beginnings of a problem and to improve proactivity. The orientations for DynamicPerf are now clear, the challenge is to accelerate the development of the work in progress on the automatic diagnosis part. Scientific objective - results and barriers to be overcome The objective is to design and validate the tooling allowing to: - correlate similar problems without established diagnosis. - make an automatic diagnosis based on a set of symptoms The data used will be generated by a DynamicPerf probe that detects unusual events on the indicators and in the logs. The main scientific and technical issues to be addressed are - Consolidation of ongoing work on symptom similarity measurement. - Development and implementation of a diagnostic method allowing clustering and classification of the observed cases. The choice of the method will be constrained by the sparsity / rarity of the events to be analyzed. - The explicability of the diagnosis is necessary to make the solution acceptable by future users in production. The main achievements or expertise expected in 1 year - Development of a demonstrator of the dynamicPerf diagnostic module implementing the proposed solution. **About you**: With a PhD, you have developed expertise in data processing and artificial intelligence. Technical skills that will be useful for the realization of the mission: - Knowledge and implementation of artificial intelligence algorithms - Software development (ideally Python, Go) - Knowledge of cloud hosting environments: VM and container infrastructures,... **Additional information**: Production monitoring tools usually use simple rules based on threshold violations, or for the latest ones, ML algorithms that allow learning and more accurate detection of anomalies. These solutions usually generate a lot of false positives and are limited to a few key indicators. The solution we propose is innovative because it is not limited to the detection of anomalies, but to their correlation in order to identify the root cause more quickly and, after diagnosis, to propose recommendations to service operators. For the diagnosis part, interesting avenues to explore have been identified in the state of the art in various fields such as computing but also in medicine or biology. **Department**: The ambition of the Innovation Division is to take Orange's innovation further and strengthen its technological leadership, by mobilising our research capabilities to fuel responsible innovation that serves human ecosystem, informs the Group's long-term strategic choices and influences the global digital ecosystem. We train the experts in today's and tomorrow's technologies and ensure continuous improvement in the performance of our services and our efficiency. The Innovation Division has 6,000 employees worldwide dedicated to research and innovation, including 740 researchers. With a global vision and a wide range of profiles (researchers, engineers, designers, developers, data scientists, sociologists, graphic designers, marketers, cybersecurity experts, etc.), the men and women of Innovation are ready to listen to and serve the countries, regions and business units to make Orange a trusted multi-service operator. **Contract**: Post Doc