Now at Alstom
I like it when software makes something real move: a train, a car, a robot.
It started with classes prépa and a dual degree in Toulon: mechatronic systems and robotics on one side, artificial intelligence on the other. In Bangkok, I worked on energy management for solar electric vehicles on a TI F28335 board. At Renault, I turned weeks of manual Excel work into Python tools that run in 30 seconds. Today, at Alstom in Pittsburgh, I build software for safety-critical railway signaling, and my work is moving more and more toward data, AI and full-stack development.
Where I’ve studied and worked
ToulonEngineering school
BangkokTFII internship
ParisRenault internship
PittsburghAlstom, today
VIEIn progress
Alstom, Pittsburgh, Pennsylvania
Photo © Alstom / C. Gandolfi
18-month VIE contract, from November 1, 2025 to April 30, 2027.
I build the tools that turn raw signaling logs into data engineers can use: a C# parser, compact files and Python dashboards. My work is moving more and more toward data, AI and full-stack software.
Certified during the VIE: DataCamp certification in Polars and Parquet.
Internal tools under NDA: no screenshots or source code can be shown.
Renault Group — Ampere Software Technology 🇫🇷
🔒 Deliverables are confidential and cannot be publicly displayed.
Six projects, one key result each. Open “How I did it” for the details, or go straight to the report or the code.
Simulation
With Ewen Guivarch
Stop a car at 100 km/h as fast as possible without locking the wheels.
4ABS controllers designed and compared
Modelled the wheel and vehicle dynamics in MATLAB/Simulink, then designed and compared four ABS controllers (continuous PI, discrete PI, on-off and an improved on-off) and added EBD front/rear brake distribution.
The improved on-off controller gave the shortest stopping distances, with wheel slip held close to the optimal 0.2. EBD prevented the rear wheels from locking early.
Data & AI
Solo project
Tell spam apart from normal text messages.
98.8%validation accuracy
Built a TensorFlow/Keras pipeline: text vectorization, word embeddings, two bidirectional LSTM layers, dropout and a tf.data input pipeline.
About 98.8% accuracy on the validation set, and all freeCodeCamp tests passed.
Robotics
With Camille Berta
Let a TurtleBot move through a cluttered room on its own.
2environments: Gazebo and the real robot
Programmed a ROS state machine in two stages: first reacting to bumper contacts (back up, turn, carry on), then processing LiDAR scans to anticipate obstacles before touching them.
The robot avoided obstacles both in simulation and in real conditions. One limit to improve: it always turned the same way.
Data & AI
Solo project
Suggest books similar to one a reader already liked.
1.1Mratings turned into recommendations
Filtered 1.1 million ratings down to active users and popular books, built a sparse book-by-user matrix and a k-nearest-neighbours model with cosine distance in scikit-learn.
Returns the five closest books for any title in the dataset, and passes the freeCodeCamp test.
Data & AI
Solo project
Beat four scripted opponents, with over 60% wins against each.
99%best win rate, 4 of 4 opponents beaten
Wrote a multi-level Markov predictor that reads the opponent’s last 4, 3 and 2 moves, falls back on move frequencies and corrects itself when its win rate drops.
99%, 78%, 65% and 63% win rates over 1,000 games against each opponent.
Data & AI
Solo project
Classify photos as cats or dogs.
72%of test images correctly classified
Trained a convolutional neural network in TensorFlow/Keras, fed by image data generators, for 15 epochs.
72% of the test images correctly classified, enough to pass the freeCodeCamp challenge.
Research internshipApril – August 2024
Thai-French Innovation Institute (TFII), King Mongkut’s University of Technology North Bangkok
I worked on the energy management of a solar electric vehicle: a control system that makes the best use of variable solar power while keeping the vehicle stable.
Everything ran on a Texas Instruments TMS320F28335 board: generating PWM signals, reading voltage and current sensors through the ADC, and driving buck and boost converters.
Ten sessions in three phases. Each one opens its PDF.
Mr. Enzo Cherif has shown exceptional dedication and skill throughout the internship, particularly in the areas of problem-solving and adaptability.
DataCampDuring my VIE at Alstom
IBMJune 23, 2025
TFII, KMUTNB, BangkokApril – August 2024
freeCodeCamp
Advanced modeling and simulation
Creating interactive applications for modeling
Design of measurement and automation applications
Robot control and coordination
Administration and development on Unix systems
Version control for collaborative projects
Scientific documents and technical reports
3D modeling and simulation for product design
Design and modeling of mechanical systems
Operating systems for everyday use
Data analysis and spreadsheet automation
Document management and professional presentations
Autonomous systems & control
Mechanical & electronic integration
Hardware & software design
Predictive models & optimization
Image processing & analysis
Connected devices & MQTT protocol
Cleaning, analysis & automation
Components & manufacturing processes
Navigation, control & safety
Lycée Parc des Loges
2019
Lycée Parc de Vilgénis
2019 – 2022
SeaTech – INP Toulon
2022 – 2025
Université de Toulon
2024 – 2025