Software Internships
|
Company
|
DELTATEC |
|
Location
|
Ans |
|
Pubication Date
|
16 Mar 2026 |
Software Internship / Master Thesis #1 – Optimization of the Video/AI Pipeline and GPU Acceleration on Nvidia Jetson Platform
Title : « Optimization of the Video/AI Pipeline and GPU Acceleration on Nvidia Jetson Platform »
Target audience : Computer Science / Bachelor, Master, Computer or Electronics Engineering
Type :
- Research and innovation 0/3
- State‑of‑the‑art analysis 1/3
- Practical implementation3/3
Supervisors : Christophe Seyler / Jean-Christophe Lallemand
Internal competence center : Embedded / PC System
Work description :
DELTATEC currently uses an embedded platform based on Nvidia Jetson, integrating a neural network for video stream processing.
In its current state, the data pipeline can be optimized to more effectively exploit the device’s hardware resources (GPU, hardware encoding/decoding units, dedicated accelerators).
The objective of the internship is to enhance and optimize the complete data path, from sensor data capture to encoded video streaming.
The work will include :
- Understanding the data flow from a video sensor.
- Efficiently routing this data to GPU resources.
- Implementing GPU pre‑processing stages before neural network inference.
- Optimizing integration with the existing neural network.
- Implementing graphical overlays (annotations, bounding boxes, contextual information) using GPU capabilities.
- Leveraging hardware acceleration for video encoding/decoding.
- Setting up an optimized pipeline enabling real-time encoded video streaming.
A key objective will be to minimize unnecessary memory copies and maximize the use of the hardware accelerations available on the platform.
Technologies involved :
- Nvidia Jetson (embedded architecture)
- GStreamer (multimedia pipeline design and optimization)
- Nvidia SDK (CUDA, possibly TensorRT, NVENC/NVDEC)
- GPU programming (CUDA / shaders)
- Real-time video processing
- Embedded Linux
Required skills
- Very good command of C and C++
- Solid knowledge of Linux (environment, debugging, compilation, system tools)
Strong assets
- Knowledge of embedded Linux
- Experience with GStreamer
- Shader programming (OpenGL)
- Basic knowledge of GPU programming (CUDA)
- Interest in image processing and embedded AI
Skills developed
- Design and optimization of real-time video pipelines
- Advanced use of embedded GPU resources
- AI integration on constrained platforms
- Memory optimization and system performance
- Full mastery of a video chain: from sensor to encoded streaming
Software Internship / Master Thesis #2 – Accelerated Multimedia Encoding
Title: « Development of a prototype enabling real-time multimedia stream encoding on GPU »
Target audience : Computer Science / Bachelor, Master, Computer Engineering
Type of work :
- Research and innovation 0/3
- State‑of‑the‑art analysis 1/3
- Practical implementation 3/3
Supervisors : Jérôme Bayaux / Stéphane Witryk
Internal competence center : Compression
Work description :
First, the trainee will study codecs, encoders, their parameters, and their impact on quality, load, and bandwidth.
Then, the trainee will build a prototype implementing all software components needed for multimedia signal encoding. The processing chain includes:
- Video essence encoding using hardware acceleration technologies (H.264/H.265 on GPU: NVENC/NVDEC…).
- Audio essence encoding.
- Multiplexing of both essences.
As the final objective is integration into a real-time chain, particular attention will be paid to constraints of the encoder, data movement management (CPU/GPU), and multithreading principles to ensure efficient and high‑performance data flow handling.
Software Internship / Master Thesis #3 – Live and Replay Streaming
Title: « Réalisation d’un prototype permettant de capturer, traiter, enregistrer et streamer un flux vidéo »
Target audience : Computer Science / Bachelor, Master, Computer Engineering
Type of work :
- Research and innovation 0/3
- State‑of‑the‑art analysis 1/3
- Practical implementation 3/3
Supervisors : Yohann Vanfrachem
Internal competence center : Cloud, Streaming, Web
Work description:
The goal is to design a system capable of processing, supervising, and historizing video streams from production lines, while significantly reducing operational complexity and cloud costs.
The solution will rely on a core principle: process as much as possible locally, send to the cloud only what is necessary, and use standardized technologies enabling precise control, smooth replay, and seamless integration of metadata produced by AI models.
The entire system is based on a unified architecture built on MPEG‑DASH for both live and replay.
An operator, via a cloud‑hosted web interface, will be able to access a slightly delayed live stream, navigate in the past using a seekable timeline, and access detection labels, statistics, and other metadata embedded directly in the video streams and archives.
Complete videos and their metadata are stored in multi-track MKV files, synchronized to the cloud only for long‑term archiving according to a controlled policy.
Software Internship / Master Thesis #4 – Multimedia Stream Receiver and Decoder App for Android/iOS
Title: « Development of a multimedia stream reception and decoding application for tablet or smartphone (Android/iOS)»
Target audience : Computer Science / Bachelor, Master
Type of work:
- Research and innovation 1/3
- State‑of‑the‑art analysis 1/3
- Practical implementation 3/3
Supervisor : Yohann Vanfrachem
Internal competence center : Multimédia Streaming / Web
Work description :
The goal is to develop a mobile application (Android preferred) capable of receiving a multimedia stream over IP (via the device’s Wi‑Fi), decoding the content (H.264 or HEVC), and displaying the result in real time.
Project details :
- Reception of the multimedia stream: Use of standard IP‑based protocols to receive the stream (typically RTSP, RTP, SRT).
- Decoding of the content: The video stream, encoded in H.264 or HEVC, will be decoded using a framework adapted to the mobile platform. The work includes researching the possible frameworks and selecting one of them (MediaCodec, ExoPlayer, FFmpeg, AVFoundation, VideoToolbox…).
- Display of the result: Development of a graphical interface using a hybrid approach: the application will embed a web page for displaying the decoded stream and handling user interactions. The work includes researching possible frameworks and selecting one of them (Flutter, React Native…).
- Development of the web page using modern web technologies: HTML, CSS, JavaScript, optionally using front‑end frameworks such as React, Vue.js, or Angular.
- Publishing of the developed application on the official store of the mobile platform.
Software Internship / Master Thesis #6 – Docker Swarm / Kubernetes / Nomad
Title : « Study and comparison of container orchestration tools»
Target audience : Bachelor in Computer Science / Master in Industrial Computer Engineering
Type of work :
- Research and innovation 0/3
- State‑of‑the‑art analysis 2/3
- Practical implementation 3/3
Supervisor : Benoît Willems
Internal competence center : IT Application
Work description :
The use and deployment of IT services through Docker containers are now widely adopted, to the point where Docker has become an essential component of any IT infrastructure.
DELTATEC, as a company developing high‑fidelity electronic and IT solutions, is no exception to this trend.
DELTATEC not only has its own IT department, but also an internal tools development team; these two services work closely together to provide the company’s engineers with a modern and reliable infrastructure supporting all engineering activities across three Business Units and an equal number of laboratories.
DELTATEC intends to expand its internal service offering, currently based mainly on a redundant cluster of virtual machines, by introducing a similar service built around Docker containers.
The expected work consists of :
- An analysis and comparison of Docker container orchestration tools (such as Kubernetes, Swarm, and Nomad) across multiple aspects: implementation, usability, flexibility, maintenance, resource consumption, reliability, etc.
- An analysis and understanding of the virtualized and containerized services currently operating within DELTATEC.
- The creation of a coherent proposal for implementing a new reliable container‑hosting service and for migrating services from the existing infrastructure to this new service.
- The development of one (or several, if necessary) laboratory proof‑of‑concepts..
Software Internship / Thesis #7 – Enhance Multimedia Tooling
Title: « Amélioration des outils d’analyse et de tests de contenu multimédia »
Target audience : Computer Science / Bachelor, Master
Type of work:
- Research and innovation 0/3
- State‑of‑the‑art analysis 1/3
- Practical implementation 3/3
Supervisors : Jérôme Bayaux / Stéphane Witryk
Internal competence center : Compression
Work description :
In the context of our graphic overlay products, multimedia content is abundant. In particular, video decoding and rendering are ubiquitous tasks that can become complex due to the large number of supported video formats.
The primary objective of the work is to improve our analysis capabilities to better assess the output obtained when decoding and rendering a given video.
The work consists in developing an application with a graphical user interface (GUI). This application will rely on independent libraries/tools (e.g., libavcodec, ffmpeg, ffprobe…) to analyze the videos provided as input. It must be capable of performing the following actions:
- Displaying generic metrics (resolution, colorimetry, codec, packing, transfer functions, required disk throughput for playback, etc.).
- Displaying Deltacast‑specific metrics: for example, detecting whether GPU‑based video decoding or pixel conversion is supported. Another example is analyzing the presentation timestamps (PTS) to verify whether they are continuous.
- Generating an analysis report in a textual format suitable for GitLab issues.
- Secondary objective to explore: simple remuxing to realign video/audio essences or even PTS.
If time allows, generation (and automation) of tests for the internal video decoding module may also be included.
Software Internship / Thesis #8 – Edge Device Vision AI
Title: « Devéloppement d’un prototype d’appareil de contrôle de processus industriel à base de vision »
Target audience : Computer Science / Bachelor, Master
Type of work:
- Research and innovation 1/3
- State‑of‑the‑art analysis 1/3
- Practical implementation 3/3
Internal competence center : Embedded / AI
Work description :
This thesis takes place within the development of a prototype device intended to evaluate the quality of an industrial process in the glass manufacturing sector.
The main goal of the internship is the development of an edge device integrating AI‑based vision capabilities (Vision AI). The system relies on the stereoscopic capture of two images of a glass plate at a specific stage of the production process. These images are then analyzed by an existing neural network responsible for detecting and identifying relevant features used for quality assessment.
The work includes:
- Analysis of the system’s technical and functional requirements
- Definition and selection of a suitable hardware platform using off‑the‑shelf components
- Integration of hardware elements (cameras, compute unit, interfaces, power supply, etc.)
- Development of startup and initialization mechanisms for the platform
- Implementation of stereoscopic image acquisition
- Implementation of the image transmission pipeline to the existing neural network
- Development of a functional prototype validating the proposed architecture
This project combines skills in industrial vision, embedded AI, hardware integration, and low‑level software development in a real industrial context. The final goal is to demonstrate the technical feasibility of an autonomous device capable of locally performing image analysis for quality control.
Prerequisites
- Basic knowledge of embedded platforms such as Raspberry Pi or equivalent
- Programming basics in C and C++
- Knowledge of Python
- Interest in embedded systems, industrial vision, and artificial intelligence
Software Internship #9 – Human Pose Estimation
Title: « State‑of‑the‑Art Review and Implementation of 2D/3D Human Pose Estimation Methods »
Target audience : Computer Engineering (Civil or Industrial) with AI orientation
Type of work :
- Research and innovation 1/3
- State‑of‑the‑art analysis 3/3
- Practical implementation 2/3
Internal competence center : AI
Work description :
Human Pose Estimation (HPE) is a key area in computer vision, with applications in robotics, sports, healthcare, animation, augmented reality, and human‑machine interaction. Recent progress in deep learning has enabled highly effective methods, both in 2D (e.g., VitPose) and 3D (e.g., SAM3D). However, these methods involve trade‑offs in accuracy, memory footprint (VRAM), inference time, ease of fine‑tuning, and portability to consumer‑grade hardware.
The goal of this internship is to conduct a comprehensive theoretical and practical analysis of state‑of‑the‑art Human Pose Estimation methods. It includes four main axes :
- Literature review of leading SOTA 2D and 3D pose estimation methods
- Implementation of the most promising methods as Proof‑of‑Concepts using a standard framework (PyTorch, TensorFlow)
- Comparative evaluation in terms of accuracy, system performance, and ease of use
- If time allows: optimization and deployment of the best solution on the target platform using ONNX and TensorRT
Software Internship #10 – Robot Framework
Title: “mproving Traceability and Readability of Automated Tests with Robot Framework»
Target audience : Computer Engineering (Civil or Industrial) with focus on AI
Type du travail :
- Research and innovation 1/3
- State‑of‑the‑art analysis 2/3
- Practical implementation 3/3
Supervisor : Charles FERIR
Internal competence center : Banc de Test, Python
Context :
Projects developed at DELTATEC are systematically accompanied by test plans describing the validation procedures for the systems.
Most tests are automated and implemented in Python to facilitate execution on test benches and improve reproducibility. However, several difficulties arise:
- Over time, test implementations may diverge from the procedures in the test plan, reducing consistency between documentation and scripts.
- Python script readability can vary greatly depending on the audience:
- developers are generally comfortable with such scripts
- reviewers or system engineers may not have the same level of mastery
This can complicate test reviews and verification of compliance.
A possible improvement is adopting Robot Framework, which provides a more declarative, keyword‑oriented approach to improve readability and maintenance.
Internship objectives :
The internship aims to assess the relevance of Robot Framework for automated testing in DELTATEC projects. The work includes:
- Learning Robot Framework and its usage principles
- Studying its integration within the current test ecosystem
- Experimenting with automated test implementation on an existing simple test bench
- Evaluating the ability to maintain better consistency between test documentation and implementation, including requirement traceability and test report generation
- Studying solutions to generate or automatically update a test plan from test scripts
- Developing keyword libraries to interface Robot Framework with existing test tools
Technologies :
- Python
- Robot Framework
- pytest
Prerequisites : Good command of PythonCompétences développées
Skills Developed :
- Design and structuring of automated test frameworks
- Improving traceability between specifications and validation
- Development of testing and automation tools
- Implementation of reusable test libraries
- Integration of test solutions in industrial environments