Customer Success - (H/F)

Senior Deep Learning Engineer – Video Surveillance
Context
Veesion develops Deep Learning algorithms for video surveillance, detecting suspicious theft-related gestures in real-time within retail environments. We work with large-scale untrimmed video data captured in real-world conditions (supermarket aisle cameras). Our mission is to turn this raw visual stream into intelligent, robust, and actionable insights.
Role
We are looking for a senior deep learning expert specialized in video who can independently design and implement state-of-the-art models. The mission is to start from scratch, with full autonomy, and outperform our current system for detecting complex human gestures associated with theft.
Requirements

• 6+ years of experience in Deep Learning, including PhD if applicable
• At least 3 years of hands-on work in video-related ML tasks (image-only experience is not sufficient)
• Strong coding skills in Python and standard DL frameworks (PyTorch, etc.)
• Comfortable working in cloud environments, and managing the full ML pipeline end-to-end
• Capable of working solo in a fast-paced, R&D-driven environment

Required Expertise
Mandatory:

• Fluent knowledge of key video datasets: Kinetics, UCF101, HMDB51, THUMOS, Something-Something, Moments in Time, Video-MME, VideoVista, LVBench...
• Strong understanding of leading video architectures: Conv3D, VideoMAE, InternVideo, Omnivec, Uniformer, ViT, Hiera, X-CLIP, Video Mamba, VideoQwen, LLAVA-Video, LongVU...
• Experience with tasks such as:
  • Video Moment Localization
  • Video Retrieval
  • Video Classification
  • Video Generation
  • Video Understanding
  • Video Instance Segmentation
  • Video LLMs

Not relevant (not sufficient):

• Standalone object tracking
• Multimodal LLMs mainly focused on image
• Traditional computer vision (image only)
• Image-only instance segmentation

Why this role is unique

• Real-world, high-impact, complex problem with major R&D opportunities
• Full technical freedom to explore and test your own ideas
• Access to large-scale proprietary video datasets and dedicated infrastructure
• Direct impact on a product deployed at scale in production