CV/Mission OpenMP freelance

Résumé des missions d'Aurèle,
freelance OPENMP résidant dans l'Ille-et-Vilaine (35)

Expérience professionnelle

QualitaX, France - Software developer
January 2025
● Development of a Minimum Viable Product: trading platform of
crypto Ethereum
Technologies: AWS DynamoDB, AWS EC2, C++, git

Huawei Technologies France, Paris - HPC/AI Research
Engineer - Freelance consultant
April 2023 - October 2024
Integrated in the SAPP research group. Communication oriented study of
Large Language Models running on Mindspore software stack and
Huawei Ascend hardware environment
● Implemented a performance benchmark of the Collective
Communication library HCCL: Measured communication
bandwidth and latency of collective primitives (Allreduce,
Allgather, ..) on Ascend hardware environment
● Distribution of LLMs on Ascend devices
● Performance analysis of LLMs’ training: Pangu Alpha, BERT,
Llama2, DeepSeek, Baichuan2, implemented in a Transformers’
library. Compared manual distribution versus automatic
distribution. Studied impact of Transformers’ features on per step
time and accuracy
● Developed a tool to extract and analyze performance metrics of
IA training workloads (per step time, communication volume,
communication throughput, computation / communication
overlap)
Technologies: Mindspore, git, Python, Bash, Huawei Ascend

Institut Mines Telecom, Palaiseau — Research Engineer
June 2020 - December 2021
● Part of the Parallel and Distributed Systems team
● Worked on the project CloudButton (*******).
○ Serverless Computing
○ FaaS: Function as a Service
● Introduced Serverless Shell: Execute shell code on a cloud
platform for data processing, using AWS Lambda, and proposed
distributed mechanisms
Technologies: Java, AWS Lambda, AWS EC2, git

Freelance, Rennes — Sabbatical leave
September 2018 - May 2020
● Skills assessment
● World trip: USA, Canada, Australia, Indonesia, Singapore

University of Oregon, Eugene, USA — Research Associate
November 2015 - August 2018
Part of the team in charge of the development of TAU Performance tool
(Tuning and Analysis Utilities)
********.php
● Developed BEACON, backplane of the Argo software stack
(*******).
● Integrated and adapted a GUI
(********-hpc/pycoolr) in order to monitor
performance metrics at application run time.
● Implemented OpenMP Tools Interface (OMPT) in the MPC
framework
● Used the MPI Tools Interface (MPI_T) to enable interaction
between TAU and MPI libraries, and inspect optimization
opportunities in the MVAPICH2 MPI library: performance
monitoring, performance recommendation and autotuning.
Technologies : C, C++, git
Publications:
1. "MPI performance engineering with the MPI tool interface: the
integration of MVAPICH and TAU" S. Ramesh, A. ********, S. Shende,
A. Malony, H. Subramoni, D. K. Panda. EuroMPI'17
2. "MPI performance engineering with the MPI tool interface: The
integration of MVAPICH and TAU". S. Ramesh, A. ********, S. Shende,
A. Malony

Exascale Computing Research, Versailles — Research Assistant
October 2011 - October 2015
Thesis subject: "Improving the Hybrid model MPI+Threads through
applications, runtimes, and performance tools"
● Collaboration with CEA/DAM
● Field: High Performance Computing
● Programming models: MPI, OpenMP
Achievements:
● Implemented an adaptive tree on the OpenMP runtime, part of
MPC, in order to lower the overhead of the OpenMP runtime on a
large spectrum of threads. Obtained a speedup of 4.9x with MPC
against the Intel Compiler, on 128 OpenMP threads.
● Optimized the MPI_Allreduce operation in a parallel manner, by
using idle OpenMP threads. Gained a speedup of 5.29x on MC, a
MPI+OpenMP application simulating transport of particles, and
relying on Monte Carlo methods.
● Unified collective operations: Implemented an optimized global
barrier as a proof of concept, involving MPI tasks and OpenMP
threads for the same operation. A speedup of 2.45x was gained
for a configuration with 2 MPI tasks and 64 OpenMP threads on a
128-core node.
● First implementation of OpenMP Tools API (OMPT) in MPC.
Technical environment: C language, git versioning tool, Linux
Publications:
1. "Adaptive OpenMP for Large NUMA Nodes". A. ********, S. Koliaï, P.
Carribault, M. Pérache, W. Jalby. IWOMP'12
2. "Optimizing Collective Operations in Hybrid Applications". A. ********,
P. Carribault, M, Pérache, W. Jalby. EuroMPI/ASIA'14

UVSQ, Versailles — Teaching Assistant
April 2013 - October 2014
Teaching Assistant (45h):
Master 1 in Computer Sciences (Versailles Saint Quentin en Yvelines
University)
- System programming: Fall 2013 & 2014
- Parallel programming (POSIX threads, OpenMP): Springs 2013, 2014

French Atomic Energy Commission (CEA),
Bruyères-le-Châtel — Research Engineer
March 2011 - August 2011
First implementation of tasks with OpenMP 3.0 inside the MPC (Multi
Processor Computing) runtime.
Technologies: C language, linux, git, OpenMP, BOTS Benchmark Suite

SCALIAN, Montrouge — Software Engineer
May 2010 - September 2010
Development on ACMS-SA, an avionic system for Sagem Defense&Security
(Airbus A350 program)
Technologies: C language, CODEO, Linux, DO-178B DAL D, PikeOS

Sogeti High Tech, Issy-les-Moulineaux — Software Engineer
December 2009 - February 2010
Design of the embedded software component ITF on the GADIRS system, for
Sagem Defense&Security
Writing pseudo code, based on input specifications
Technical environment: C language, Rational Test Real Time, DOORS, norme
DO-178B niveau A
● Integration and validation of the embedded driver NAVPLD on the
GADIRS system, for Sagem Defense&Security
● Writing cases tests, and scripts, based on input specifications
● Validate tests following D0-178B standard
Technical environment : C language, Rational Test Real Time, DOORS,
Synergy, norme DO-178B DAL A, Arinc 429
Development of an exercice editor, in order to simulate air traffic equipments
for Thales Air Systems
● Development of GUIs
● Import and export of XML data
Technologies: Java, Swing, JDBC, XML

Thales Services, Vélizy — Developer
February 2008 - August 2008
Developed, tested and integrated web services connected to a supervision
system, for metro of Brussels
- Development of unit tests
Technologies: J2EE, Axis 2, Serveur Apache/Tomcat

Thales Optronics SA, Guyancourt — Developer
April 2007 - September 2007
Integrated a web service in a CORBA environment for the naval infrared
system ARTEMIS
- Debug and port of the web service on CORBA environment
Technologies: C language, Linux Red Hat, CORBA