INSIGHTS

Machine Learning Engineer – Computer Vision & Video Models

Machine Learning Engineer – Computer Vision & Video Models

You build and train the vision models behind a real-time quality-control platform for one of the world’s largest live casino providers. You work closely with the Lead Computer Vision Engineer and turn studio footage into models that hold up in production.

About K&C

K&C (Krusche & Company) is a Munich-based technology partner for medium and large European companies. Our senior engineering teams in Poland, Ukraine, Bulgaria and Egypt build, modernize and scale business-critical software for clients such as Nestlé, Bosch, Sparkassen and Informa.

Our promise is Beyond Outsourcing®: deep technical ownership, German business standards and engineering excellence. K&C AI Solutions brings this to computer vision, machine learning and AI transformation.

The project

Live casino studios stream dozens of tables to players worldwide, 24/7. Every round follows precise dealer procedures for card placement, pulling technique, hand gestures and card handling. The platform checks these procedures automatically and flags every deviation to the quality team.

  • 50 concurrent image and video streams
  • Detection accuracy of at least 85% across procedure categories
  • No more than three models in production, combined with a rule engine
  • Fully annotated training data provided by the client as images (JPG, PNG) and video (MP4)
  • Delivery starts with card and placement checks and expands to hand and gesture recognition

What you will do

  • Train and fine-tune object detection, hand-keypoint and temporal action recognition models (e.g. RT-DETR, RTMPose, skeleton-based action recognition) on the client’s studio footage
  • Build data loaders and preprocessing pipelines for images and video, including frame sampling, augmentation and class balancing
  • Tackle rare events and imbalanced classes such as hidden cards, covered logos and incorrect action timing
  • Build the training and evaluation framework: metrics per procedure category, backtesting and validation reports
  • Run error analysis and turn findings into annotation and data quality recommendations
  • Implement detection logic in the rule engine that combines model outputs into clear findings
  • Prepare models for production: export to ONNX, quantization and TensorRT optimization together with MLOps
  • Contribute to model versioning, Model Cards and technical documentation
  • Take part in knowledge transfer sessions with the client’s engineering team

What you bring

  • 3+ years in machine learning, with at least 2 years in computer vision
  • Hands-on experience training detection, pose or keypoint models on real-world image or video data
  • Experience with video understanding, action recognition or temporal modelling
  • Strong Python and PyTorch; working knowledge of ONNX and model optimization
  • A rigorous approach to evaluation: precision and recall trade-offs, validation splits, avoiding leakage between video sequences
  • Experience with imbalanced datasets and rare-event detection
  • Familiarity with experiment tracking (MLflow, Weights & Biases or similar)
  • Fluent English and the ability to explain model results clearly to non-ML colleagues

Strong plus

  • Hand-object interaction, gesture or sign language recognition
  • Industrial step-compliance, surgical video or sports pose-tracking projects
  • TensorRT, Triton or NVIDIA DeepStream
  • Annotation tools such as CVAT
  • Live casino or iGaming background

Send us your CV and Cover Letter

"*" indicates required fields

Full Name*
Accepted file types: pdf, doc, docx, Max. file size: 5 MB.
Data Protection*
Data Processing