Functional Modeling of a DNN Accelerator H/F (Sciences pour l'ingénieur)

Job description

Domaine : Sciences pour l'ingénieur

Contrat : Stage

Description du poste :

As part of the ongoing research on artificial intelligence at the CEA, a hardware architecture for a Deep Neural Network accelerator (DNN) has been developed. The purpose of this internship is to develop a functional model of the accelerator using either SystemC or Python.
The student will join a development team and will write a model including the core of the accelerator and then the integration of the accelerator into the complete system. Part of the work will include contributing to the micro-architecture and data-flow in the machine, in order to optimize both the compute performance and energy efficiency. Depending on how quickly the model is developed, he/she may also be asked to port specific AI algorithms to execute on the model.
This internship will take place in the AI team within the LSTA laboratory which is part of the DSCIN department at the CEA LIST in Grenoble.
The student will have access to a full CAD environment to develop and verify their model. They will have the opportunity to interact with the software and hardware design teams and will participate in technical meetings.
Throughout the development effort, the student will be responsible for evaluating the performance of their model using a set of standard benchmarks (either existing tests or new tests which will be developed).


Student in their final year of an engineering program or in a masters program specialised in computer-science, computer architecture or micro-electronics
Desired Skills:
 Programming in C/C++, System C or Python
 Knowledge of AI algorithm principles
 Basic knowledge of processor architecture
 Ability to work in a team

Ville : Grenoble