Yoann Roth Email & Phone Number

Cambridge, Gb

Cambridge, England, United Kingdom

Part of the GPU and Deep Learning team. The team's responsibilities includes: Deep Learning performance, GPU integration for Deep Learning and GPU platform support.Responsibilities included - Take part of the Deep Learning Toolbox development effort which includes bug fixes, code review, design and implement features using the RFAIN process - While the team is managed by a Senior Team Lead, act as Team Lead to assist their role and ramp up on managerial responsibilities - Attend and represent the company at various customer facing event, for example NVIDIA GTC in March 2024 - Supervise projects done by engineers from MathWorks graduate programTeam Lead activities (since May 2022) - Organize planning and strategy for the Deep Learning Performance part of the team's responsibility, under the supervision of my manager - Take-on managerial responsibilities with a team member, doing 1-on-1, planning their tasks and professional development.Examples of development for Deep Learning Toolbox - Implemented multiple large scale refactors of the MATLAB codebase to improve performance and extend flexibility for future features oriented towards performance for Deep Learning (MATLAB) - Contributed to the architecture implementation of the new trainnet Deep Learning training function and implemented its support for parallel training (MATLAB) - Maintained integration with NVIDIA C++ cuDNN library and prototyped its migration to the cudnn_frontend library, including the scaled dot product operation used in the attention layer (C++) - Prototyped integration with NVIDIA TensorRT library using their C++ API (C++)Examples of development for Parallel Computing Toolbox - Integrated NVIDIA CUDA 11.8 in MathWorks products, a six months efforts requiring coordinating multiple engineers (MATLAB, C++, Makefile) - Developed a internal prototype for a GPU monitor, reporting GPU activity to the user via a User Interface (Javascript React, C++).

Cambridge, United Kingdom

Part of the GPU and Deep Learning team. The team's responsibilities includes: Deep Learning performance, GPU integration for Deep Learning and GPU platform support.Responsibilities included - Fixing bug, designing and implementing features using the RFAIN process - Provide feedback to other developers via design reviews and code reviewsDevelopment for Deep Learning Toolbox - Contributed to the new framework released in R2019b using the dlarray object, implementing some of the deep learning specific methods - Added some data management features such as minibatchqueue - Added and maintained various optimizations applied during propagation in a Deep Learning network - Maintained integration with Parallel Computing Toolbox for features such as parallel training or DispatchInBackground - Maintained integration with NVIDIA C++ cuDNN library

Cambridge, United Kingdom

This role is part of the Engineering Development Group (EDG, MathWorks Graduate program)Responsibilities included - Assisting customers with technical issues, answering emails and phone requests - Acting as Group Leader (on a rotation basis) to triage issues to the team - Assisting managers in preparing the team's planning - Becoming familiar with a large number of MathWorks products - Undertaking development projects with the dev teams (4 projects total)

Grenoble, France

PhD Thesis: "The Physical Layer for Low Power Wide Area Networks: A Study of Combined Modulation and Coding Associated with an Iterative Receiver".- Investigated modulation and coding schemes for long range applications;- Developed an innovative solution based on an iterative receiver, to be integrated in an industrial project;- Achieved high technical expertise in telecommunication systems;- Published over 5 international papers and 1 patent (filled).Director : Laurent Ros (laurent.ros@gispa-lab.grenoble-inp.fr)Supervisor : Jean-Baptiste Doré (jean-baptiste.dore@cea.fr)Material (thesis and defence presentation) are available at https://hal.archives-ouvertes.fr/tel-01568794/

Grenoble, France

Master Thesis: "4G transceiver: RF modelling and DC oset compensation".- Designed a DC offset cancellation algorithm for the company’s LTE receiver;- Implemented the algorithms in Matlab and C++ simulators;- Studied telecommunication standards (LTE, UMTS) and signal processing algorithms.Supervisor : Fabrice Belvèze (fabrice.belveze@st.com)

Grenoble, France

Bachelor Thesis: "Faster-than-Nyquist transmission"- Conducted a bibliographic study on the Faster-than-Nyquist- Implemented Matlab model - Wrote a technical reportSupervisor : Cyrille Siclet (cyrille.siclet@gipsa-lab.grenoble-inp.fr)

Doctor Of Philosophy (Phd), Signal And Telecommunications

Activities and Societies: Junior Entreprise (innovDoc), Web Chair of the Crowncom2016 conference

Master Of Engineering (M.Eng.), Signal Processing And Telecommunications

Activities and Societies: Membre de l'association du bureau des élèves de Phelma 2011-2012