About Me
Sylvain is an highly experienced R&D engineer with over 30 years of international experience. His skills range from AI model quantization, to optical spectrum analyzer design to mobile communications and telephony design. Sylvain’s expertise lies in both hardware and software DSP, algorithms, and Machine Learning optimization.
Experiences
- Machine Learning Consultant
- 2021 - Present · SHD Group
- Machine Learning, Algorithms, DSP
- Principal Engineer
- 2020 · 0JAS
- Machine Learning Engineer
- Principal Engineer
- 2017 - 2019 · Wave Computing
- Team lead, Core Technology, AI group
- Senior Staff Engineer
- 2015 - 2016 · Qualcomm
- Engineer in the Firmware group, responsible for algorithm designs, simulation, modelisation, and implementation in the area of VDSL.
- Principal System Engineer
- 2003 - 2016 · ikanos
- Principal engineer in the system-algorithm team.
- Senior Design Engineer
- 1984 - 2003 · Anritsu (formerly Wiltron)
- Hardware engineer. Board design, PLD, FPGA, DSP, telephony.
Education
- MSEE, focus on Telecom and DSP
- 1983 - 1984 · Stanford University
- Diplôme d'Ingenieur (5-year degree), Mechanical Engineering, Linear Systems
- 1978 - 1981 · ENSMM - École Nationale Supérieure de Mécanique et des Microtechniques
- DEA (MAS), Optics and Signal Processing
- 1980 - 1981 · Université de Franche-Comté
- Classes Préparatoires
- 1976 - 1978 · Lycée Pierre-de-Fermat, Toulouse, France
Patents
Patents
2019. Flamant, Sylvain. Neural Network Output Layer for Machine Learning. U.S. Patent Application Number 16/459,731, Patent Pending.
2015. Flamant, Sylvain et al. Method and Apparatus for Peak-to-Average Ratio Reduction. U.S. Patent 9,036,730 B2.
Recent Projects
-
Efficient Softmax
- Designed a 16-bit fixed point softmax layer for hardware implementation using 3 small lookup tables in order to only use 2 multiplications per output.
-
Efficient SqueezeNet
- Simulated the SqueezeNet NN for low number of bits (8-16 bits) for inference and training.
-
Simple Batchnorm
- Designed a simpler batchnorm block with equivalent performance for hardware implementation.
Skills
Machine Learning
PyTorch
Neural Network Simulations
DSP
Algorithms
Hardware
Firmware
Python
C
Mathcad
MATLAB
Hobbies
- USA Swimming Official
- Swimming
- Hiking
- Traveling