🤓

Gaetan Bahl, PhD

Principal AI/ML Engineer

NXP Semiconductors

I am a ML Engineer with 8+ years of experience developing deep learning solutions for embedded systems. I joined NXP in 2022 after finishing my PhD on Deep Learning architectures for onboard satellite image analysis, under the supervision of Florent Lafarge at Inria. My expertise includes computer vision algorithm optimization, neural network compression, and deployment on resource-constrained devices.

This website is hosted on a Raspberry Pi.

Recent Publications

PhraseStereo: The First Open-Vocabulary Stereo Image Segmentation Dataset

Introduction of PhraseStereo, the first open-vocabulary stereo image segmentation dataset, presented at ICCV 2025 X-Sense Workshop.

Thomas Campagnolo, Ezio Malis, Philippe Martinet, Gaetan Bahl

Single-Shot End-to-end Road Graph Extraction

We propose a method that combines a Fully Convolutional Network and a Graph Neural Network to efficiently extract road networks from aerial images in a single pass.

Gaetan Bahl, Mehdi Bahri, Florent Lafarge

Single-Shot End-to-end Road Graph Extraction

SCR: Smooth Contour Regression with Geometric Priors

This paper presents a method that captures resolution-free object contours as complex periodic functions.

Gaetan Bahl, Lionel Daniel, Florent Lafarge

SCR: Smooth Contour Regression with Geometric Priors

Scanner Neural Network for On-board Segmentation of Satellite Images

Using 1D convolutional LSTM to efficiently process satellite images of arbitrary length.

Gaetan Bahl, Florent Lafarge

Binary Graph Neural Networks

We present and evaluate different strategies for the binarization of graph neural networks. We show that through careful design of the models, and control of the training process, binary graph neural networks can be trained at only a moderate cost in accuracy.

Mehdi Bahri, Gaetan Bahl, Stephanos Zafeiriou

See all publications

Projects

TikZ Roadmap Visualization

A LaTeX package for creating beautiful roadmap visualizations using TikZ, designed for academic papers and presentations.

TinyPysics

A lightweight Python physics engine for educational purposes and simple simulations.

NXP eIQ Model Zoo

Collection of optimized deep learning models for NXP’s eIQ machine learning software, featuring 30+ models across various computer vision tasks.

Cloud segmentation and Boat detection on Jetson Nano and Raspberry Pi at AzurIA

Training lightweight convolutional neural networks for real-time embedded processing of aerial images.

Cr'Hope (Hope for Crop)

Insect pest detection and recognition, with simulation of spread on edge device.

Airbus Pixel Factory Neo

Pixel Factory suite represents a state-of-the-art industrial geo-processing system designed to produce a wide range of cartographic end product of unparalleled quality.

CARAPAS surgery simulator

Took part in the creation of a patient-specific surgery simulator based on CT scans and a biomechanical model of the hip, using Artisynth, a physics simulation framework in Java.

BPM detection using Wavelets

Detect time-based rhythmic features and retrieve the tempo of the song.

Fast image segmentation using Level Sets

I implemented a level set algorithm from Yan Wang and Chuanjiang He and used it for image segmentation.

Flocking simulation using SFML, OpenMP and OpenCL

Bird-oids, or Boids, are flying in flocks and try to avoid the mouse pointer. The goal is to have as many boids as possible while maintaining an acceptable framerate.

Reaction-Diffusion model for population simulation

We chose to work on a 2D simulation of the evolution of a population using a Reaction-Diffusion model.

Simulation of the air-cooling of electric components

We simulate the cooling of a PCB and two CPU’s using Finite Elements in the FreeFem++ software.

Dice Recognition in OpenCV using K-Nearest Neighbor algorithm and Support Vector Machines

The goal is to read numbers on dice and compute the score using OpenCV and the K-Nearest Neighbor algorithm or Support Vector Machines.

Ensimag arcade machine

Fixing and modifying an old arcade machine to play SEGA Genesis games

Music genre recognition using Fourier Transform and Wavelets

The goal is to recognize the genre of a music piece using features extracted from the FFT and DWT.

Contribution to ARM Compute Library

Implementation of computer vision algorithms in ARM NEON during my internship at ARM.

Introduction to Lab Research

Viscosity solutions and associated numerical schemes

Network optimization using a bio-inspired algorithm

Studying the behaviour of Physarum polycephalum, a slime mold that is able to find its way through labyrinths as well as build short and robust networks to feed itself.

Guitar string simulation

Modelling and simulation of a guitar string and an eardrum in Scilab using finite differences

JPEG encoder and decoder in C

Professional Experience

Principal Machine Vision Engineer

NXP Semiconductors

January 2025 – Present Sophia Antipolis, France

Responsibilities include:

Leading Vision Models and Research team (3 FTEs + 1 PhD candidate + interns)
Setting research direction for embedded computer vision and robotics applications
Supervising industrial PhD candidate on robotics applications in partnership with Inria ACENTAURI
Mentoring junior engineers and managing academic partnerships
Public outreach and conference reporting
Contributing to CES 2026 face recognition demo

Senior Machine Vision Engineer

NXP Semiconductors

May 2022 – December 2024 Sophia Antipolis, France

Responsibilities include:

Research and implementation of deep learning methods for embedded vision
Stereo depth estimation, neural network compression (quantization, pruning)
Development of vision pipelines for NXP i.MX applications processors and microcontrollers
Created NXP eIQ Vision Model Zoo with 6 releases and 30+ optimized models
Contributed SIMD optimizations to GStreamer ORC library
Designed and delivered nationwide ML training program, upskilling 100+ NXP engineers
Technical evaluation of Generative AI-based tools for software engineers

Machine Learning Engineer

AzurIA

February 2022 – May 2022 Sophia Antipolis, France

Embedded computer vision (boat detection, cloud segmentation, anomaly detection) on Jetson Nano and Raspberry Pi. Demo.

PhD Student

Inria / IRT Saint-Exupéry

February 2019 – February 2022 Sophia Antipolis, France

Deep Learning architectures for onboard satellite image analysis.

Software Engineer

Airbus Defense and Space / Meritis

October 2017 – February 2019 Sophia Antipolis, France

Development of an aerial image processing pipeline for ortho-image and digital terrain model production on Google Cloud. Semantic segmentation of 3D city models.

Software Engineering Intern

TIMC-IMAG Lab

February 2017 – October 2017 Grenoble, France

Worked on a patient-specific surgery simulator in Java using a physics simulation framework (Artisynth)

Software Engineering Intern

ARM

June 2016 – September 2016 Cambridge, United Kingdom

Computer Vision and Image Processing Internship

Software Engineering Intern

CEA

June 2015 – August 2015 Grenoble, France

Created and optimised an R package for the analysis of High Throughput Screens.

Skills

Deep Learning

TensorFlow/Keras, PyTorch, Model Optimization

Computer Vision

Stereo Vision, OpenCV, GStreamer, Semantic Segmentation

Programming

Python, C/C++, Go, Java, Bash

Embedded Systems

ARM, NXP i.MX, FreeRTOS, ROS2

Cloud & Tools

Google Cloud, Docker, Git, Linux

Other activities

Teaching Deep Learning at Polytech’ Nice and EIT Digital Master
Teaching Data Science at Polytech’ Nice (2019-2022)
Reviewer: TVCJ, IJCV, ISPRS Ph&RS, ESWA, CVPR, EARTHVISION, IGARSS, ECCV, WACV, …
Founder and administrator of NXP Semiconductors Sophia Antipolis board game club (since 2023)
Hobbies: Photography, Music (Piano, Trombone, Computer-assisted music), Videomaking, Gaming, Motorcycling, Aviation

Gallery