Emrick Sinitambirivoutin

I graduated from Grenoble INP ENSIMAG with a major in embedded systems and software devices. In addition to that major, I studied machine learning and data science during an academic exchange at the Ecole polytechnique fédérale de Lausanne (EPFL) in Lausanne.

Always eager to put into practice my skills into real projects, I did a lot of interships during my studies. Here are some examples of the projects I’ve worked on during those:

• Convolutional neural Network accelearation on FPGA: During my first 6 months internship, I worked at Atos R&D center at Echirolles (near Grenoble). I studied the implementation of an FPGA accelerator for inference with my tutor Alban Bourge. I was mainly focused on the design of an automatic network generator flow using Python. From a CNN stored in ONNX format, the tool was able to generate an optimized hardware architecture of the network on the FPGA from generic blocks. The solution achieved state of the art results on some simple CNN architecture. I co-authored a patent on our solution with the team

• Optimizing heterogeneous acceleration pipeline: During my second internship, I worked at IBM Power Systems lab in Montpellier. I worked on the design and the implementation of massive data acquisition and processing pipeline using FPGAs and GPUs on Power9 servers. I was focused on the optimization of data exchanges beteen a FPGA and a GPU using IBM Coherent Accelerator Processor Interface (CAPI) and NVIDIA NVLINK technology to maximize throughtputs in the pipeline. I made some open-source contributions to the SNAP framework developped by IBM teams and available on GitHub

I’m now a Machine Learning Engineer at Sonos where I work on On Device Speech Recognition.