Freelancer profile translated to English.
Back to original languageAbout Maxime
ML Engineer — Video Intelligence & Multimodal AI. I take your AI projects from R&D to production, at scale.
5+ yearsin video intelligence and multimodal AI. My specialty: bringing an AI project **from POC to production**, and running it at scale at the right cost.
At Canal+,I industrialize R&D systemsfor the global catalog: large-scale video indexing (15k hours/run), multimodal agent search (real-time chat + voice), LLM/VLM pipelines. I don't deliver a notebook that works once — I deliver ascalable, monitored, and cost-optimized architecture.
Across the entire chain:
-R&D & scoping— from a vague need to a well-defined ML problem, model benchmarking (LLM, VLM, CV), rapid feasibility proof.
-Prototyping— a demonstrable POC, quickly: semantic search, conversational agents, detection/segmentation/tracking models.
-Industrialization & MLOps— automated pipelines (AWS Step Functions, Lambda, Docker), APIs (FastAPI), CI/CD, scalable workers.
-Production & scaling— lean and scalable architecture: indexing tens of thousands of hours, multi-index search, FAISS at 200 ms on millions of records.
Stack:Claude Agent SDK, OpenAI, Gemini, RAG, full-duplex voice agents · PyTorch, YOLO, U-Net, OpenCV · OpenSearch, FAISS, vector DBs · AWS, Docker, CI/CD, FastAPI.
Rigorous, product-oriented, obsessed with delivering at scale and at the right cost. I support you throughout the cycle — or just on the link you're missing.
Available for your video intelligence, multimodal AI, LLM agent, and computer vision projects — from the first prototype to production.
French
Native or bilingual
English
Native or bilingual
Can work on-site
Paris (up to 50km)
Experience
- Canal+Machine Learning EngineerFILM AND AVSeptember 2023 - Today (2 years and 11 months)Paris, FranceDevelopment of a scalable pipeline using a VLM for the "Et ta Scène?" feature, reducing costs by 90%- Design of the "Et ta Pub?" system for automated chapter extraction and ad placement optimization using LLMs- Creation of "Et ta Marque?" for brand recognition in movies/series using LLM technology- Design of "Et ton Thème?" for intelligent content chaptering with subject, context, and viewpoint analysis- Implementation of metrics to evaluate content diversity and plurality- Automation of workflows via AWS Step Functions and Lambda- Development of generic workers for in-depth analysis and understanding of global content
- JakartoComputer Vision EngineerAUTOMOBILEJune 2023 - August 2023 (3 months)Montréal, CanadaClustering approach for traffic signs using deep learning:- Development of an optimization module to improve object separation within clusters, thereby enhancing the accuracy of object classification in urban inventory.- Utilization of advanced data capture systems, including LiDAR scanners, cameras, and high-precision GPS for triangulation and georeferencing.- Use of a Siamese network to train an encoder that facilitates object separation by the clustering algorithm.- Conducted comprehensive experiments using the Fashion MNIST and traffic sign datasets to evaluate the performance of the developed module.- Achieved a v-measure of 0.94, demonstrating the high efficiency of the developed process for differentiating objects within a cluster.- Contributed to improving the efficiency of urban inventory by enabling more accurate classification of detected objects.
- Hache de LancerComputer Vision EngineerSPORTSMay 2023 - August 2023 (3 months)Paris, FranceDesign of vinyl record cover recognition for an application:- R&D feature engineering for recognition (200 ms end-to-end inference request)- Identification of the nearest vector in a large dataset within the FAISS index (millions of vinyls)- GitHub CI/CD and deployment on Scaleway- GitHub Actions
Reviews
Recommendations
FU
FU
+1
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
- AI Engineer and Big DataTELECOM NancyUne formation « généraliste » développée à l’école au travers de cinq axes : trois axes couvrant le « cycle de vie d’un produit, logiciel ou service informatique » et deux axes transverses, couvrant les sciences fondamentales et appliquées ainsi que celles du management, de la communication et de l’innovation. L'approfondissement « Intelligence Artificielle et masses de données » est développé en entreprise (LogIC SAS).
- CPGE TSILycée Louis VincentClasse préparatoire aux grandes écoles en technologies et sciences industrielles. Approfondissement en Mathématiques, Physique et Science de l'ingénieur.