Freelancer profile translated to English.
Back to original languageAbout Moctar
Computer Vision & Deep Learning/Embedded AI Specialist (Python, C/C++)
Looking for a Computer Vision/AI Engineer who can transform your R&D POCs into reliable and optimized products?
You've come to the right place.
With over 10 years of experience in Image Processing, Computer Vision, and Deep Learning for embedded systems, I excel at bridging the gap between research and production, optimizing your solutions for real-time performance.
Image Processing, Computer Vision, Deep Learning, LLM, GPU computing.
Concrete Achievements
Design and deployment of algorithmic solutions to improve the image quality of GoPro cameras (classic image processing algorithms and neural networks optimized for embedded systems) for consumer use.
- Design and deployment of algorithmic solutions (image processing, Computer Vision, and CNN) for real-time license plate localization and reading.
- Design of a highly optimized OCR for DSP (C implementation of CNN inference for OCR).
- Design and deployment of a monitoring and real-time anomaly detection module (Cropping, Defocus, etc.) for Survision cameras.
R&D on complex image processing and automatic interpretation problems for seismic data.
Tailored Solutions
- Analysis of your needs.
- Transition from POC (Python) to production (C/C++ and CUDA).
- Optimization of your solutions for real-time performance.
Contact me to transform your Computer Vision projects into robust and high-performance solutions.
French
Native or bilingual
English
Fluent
Arabic
Basic
Yoruba
Native or bilingual
German
Basic
Remote only
Primarily works remotely
Experience
- ThalesPython/C++ Developer, Video Processing - Low LatencyDEFENSE AND MILITARYApril 2025 - Today (1 year and 2 months)Osny, FranceReal-time optimization & multithreaded architecture- Design and implementation of a high-performance multithreaded architecture for parallel processing of video streams (acquisition, decoding, processing, rendering).- Intensive use of synchronization primitives (mutex, atomic, condition_variable), optimization of task partitioning to minimize blocking and cross-thread contention.- Implementation of a fully asynchronous real-time pipeline to reduce latency between stream capture and display.- Measured reduction of overall pipeline latency by over 30% after redesign.- Development of thread-safe C++ algorithms for simultaneous management of multiple HD and 4K video streams, with strict throughput and latency constraints.CPU/Memory Optimization- Detailed performance analysis via CPU/GPU profiling (perf, Valgrind, Nvidia Nsight) to identify bottlenecks in decoding, conversion, and memory transfer stages.- Complete rewrite of several critical modules to:- improve CPU cache utilization (cache-friendly structures),- reduce dynamic allocations,- limit memory copies (zero-copy, circular buffers),- Optimize frame processing time across the entire pipeline:- Reduction of locks,- Implementation of lock-free queues for certain pipeline sections.Project: Video Streams- C++ development of secure, real-time video stream transfer algorithms- C++ development of video conversion algorithms to various formats- Development of a REST API server for monitoring video recording and replay operations;- Optimization of real-time video stream transfer pipelines: redesign of transfer mechanisms, storage data structures, data decoding, and display. Multithreading;- C++, Python, OpenGL, Cuda, REST API, FLASK, Multithreading, CI/CD / tests
- SurvisionR&D Engineer in Embedded Image ProcessingTECHSeptember 2019 - April 2022 (2 years and 8 months)Nanterre, France• Development of embedded image processing algorithms for automatic, real-time license plate recognition (LAPI)• Proposal and training of CNN neural networks for license plate localization and automatic character recognition (OCR)• TinyML: Lightening and optimizing neural network architectures for plate localization and OCR for embedding in LAPI cameras• C implementation and optimization of CNN inference for OCR to be embedded in an old camera with a C6x DSP: data rearrangement to minimize cache misses, reduce data transfer and computation times• Development and integration of image processing and/or machine learning algorithms for:✓ Autofocus for a new range of cameras with a variable zoom✓ Real-time anomaly detection on LAPI cameras (Cropping, blur, etc.) that could reduce camera performance.
- GoProSenior Engineer in Embedded Computer Vision and Deep LearningTECHJune 2022 - March 2025 (2 years and 9 months)Issy-les-Moulineaux, France- Participation in the creation of an AI module integrated into GoPro cameras to improve the image quality of photos taken with GoPro cameras• Deep Learning: Design and training of neural networks for real-time image restoration (denoising, super-resolution) for GoPro cameras.• TinyML: Optimization and lightening of proposed neural network architectures (distillation, pruning, quantization, etc.).• Deployment of the neural network for image denoising in GoPro cameras• Design and training of Diffusion Models for image restoration, NLP, LLM.- Development of algorithms, optimized for embedded systems, to improve the ISP and thus the image quality in GoPro cameras (Demosaicking, Sharpening, Gamma Correction, etc.);- Optimization of Stitching algorithms for the 360 camera,- Geometric and vignetting calibration, etc.- Development of non-regression tests for code monitoring (Unittest, Github actions)• Python, TensorFlow, Keras, OpenCV, Bash scripting, Git, Github, CI/CD
Recommendations
SN
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- PhD in Image ProcessingUniversities of Bordeaux2018
- Electronics Engineer, Specializing in Signal and Image ProcessingENSEIRB-MATMECA of Bordeaux2014
Certifications
- Machine Learning in Python with Scikit-learnFrance Universités Numérique2022
- Neural Networks and Deep learningDeepLearning.AI2022