2022-05591 - Ultra Low-power AI for Embedded Devices
Le descriptif de l’offre ci-dessous est en Anglais

Contrat renouvelable : Oui

Niveau de diplôme exigé : Bac + 5 ou équivalent

Autre diplôme apprécié : Masters of Science + experience

Fonction : Ingénieur scientifique contractuel

Niveau d'expérience souhaité : De 3 à 5 ans

Contexte et atouts du poste

The context of this position is a partnership with Orange and Freie Universität Berlin (TinyPART project) as well as with the RIOT community, around the topic of experimental low-power AI and TinyML. The goal is to implement multi-scale Machine Learning, applicable not only in the cloud and at the edge but also on microcontrollers.

Mission confiée

Missions :
Support, evaluate and develop experimental Machine Learning software libraries for low-power connected objects. Set up demonstrators using Machine Learning in the context of the thing-edge-cloud continuum.

For a better grasp on the targeted topics :
See practical literature such as TinyML [1] for inference on microcontroller, as well as some of the recent existing research articles on learning on microcontroller (such as [2][3]), low-power federated learning (such as [4]), and keynotes on the topic such as [5]. Also see embedded ML software bases such as those proposed by EdgeImpulse, TensorFlowlite-micro, etc., and embedded software platforms for 32-bit microcontrollers, especially RIOT [6].).

References:
[1] T. Warden, D. Situnayake, "TinyML", O'Reilly, 2019..
[2] MCUnetv3 https://tinytraining.mit.edu/
[3] H. Ren et al. "TinyML with Online-Learning on Microcontrollers" Proceedings of IJCNN, 2021
[4] MM Grau et al. "On-Device Training of Machine Learning Models on Microcontrollers With a Look at Federated Learning." ACM GoodIT, 2021.
[5] C. Adjih, "Machine Learning for IoT", Workshop on IoT and Emerging Technologies, 2022.
[6] RIOT operating system for low-power IoT.

Principales activités

Main activities :
- Review state of the art for on-device learning on microcontrollers
- Test/benchmark existing solutions on supported low-power hardware
- Integrate selected solutions in RIOT (aiming to extend and facilitate wider hardware support)
- Prototyping and demonstrations with low-power TinyML (e.g., HCI gesture detection voice command...)
- Contribute to modify/develop engine for dynamic execution of inference models


Complementary activities:
- Upstreaming of code in the RIOT ecosystem and implementation of CI
- Organization of hackathons
- Academic experimental research publications & documentation in the field of TinyML

 

Collaboration :
Some trips/stays in Berlin may be realised in this context, thanks to our collaboration with Freie Universität Berlin on this topic.

Compétences

Technical Skills and Level Required:
- Knowledge and proficiency in Machine Learning techniques
- Knowledge and mastery of Python, and low-level C programming language and tools
- Knowledge and mastery of low-level software optimization techniques, e.g. on 32-bit microcontrollers


Languages :
- Good command of scientific English;


Interpersonal skills :
- Teamwork (partially geographically distributed).

Avantages

  • Subsidized meals
  • Partial reimbursement of public transport costs
  • Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
  • Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
  • Professional equipment available (videoconferencing, loan of computer equipment, etc.)
  • Social, cultural and sports events and activities
  • Access to vocational training
  • Social security coverage