Senior Computer Vision Engineer - Foundation Models (m/f/d)

Are you passionate about advancing the field of Computer Vision and eager to apply your expertise to impactful, real-world problems? We are seeking a highly skilled Senior Computer Vision Engineer to join our dynamic team and play a pivotal role in developing our proprietary foundation models.

Your New Passion:

• Develop and train domain-specific vision models using state-of-the-art architectures and our unique waste image dataset for pretraining and continued pretraining.
• Design and manage our end-to-end ML training pipeline, from data preparation and distributed training (PyTorch FSDP/DDP, Mixed Precision) to model versioning.
• Build and maintain our evaluation suite, the core infrastructure for measuring model performance, including linear probing, k-NN probing, few-shot detection, cross-domain generalization, and anomaly detection.
• Fine-tune and distill models for specific downstream tasks and edge hardware (sorting systems, GPU inference).
• Systematically analyze training runs, identify issues like feature collapse or domain shift, and develop sustainable solutions.
• Collaborate closely with the Cloud Backend team to efficiently deploy models (ONNX, TensorRT, OpenVINO).
• Actively track research advancements in Computer Vision and translate relevant papers into productive solutions.
• Consider the broader impact of your work on real-world operations, sorting systems, customers, and the overall system.

What Excites Us:

• Several years of experience in developing and training Computer Vision models, ideally with modern Vision Transformer architectures and Self-Supervised Learning methods.
• Proficiency in PyTorch, including distributed training (DDP, FSDP), mixed precision (bf16/fp16), and performance optimization (torch.compile, profiling).
• A deep understanding of model training and evaluation, recognizing the critical importance of a robust evaluation suite.
• Experience with modern ML tooling stacks for configuration management, experiment tracking, data versioning, and core libraries.
• Proficiency in using AI tools (e.g., Claude, Copilot) to accelerate routine coding and focus on complex research and architecture challenges.
• A strong grasp of data pipelines for large datasets (millions of images), including efficient data formats, GPU augmentations, and I/O bottlenecks.
• Experience with common detection/segmentation and anomaly detection frameworks is a plus.
• Knowledge of inference optimization and model distillation, with prior experience deploying models on edge hardware being ideal.
• Exceptional problem-solving skills, analytical thinking, and scientific rigor; a hypothesis-driven approach.
• Comfort with cloud GPU infrastructure (AWS, Azure, GCP, or on-premise H100/A100 clusters).
• Fluent German and good English language skills.
• Ideally, research experience (papers, open-source contributions, conference talks) or a PhD – not mandatory, but a plus.

What You Can Look Forward To:

• A 'greenfield' development environment, building a foundation model strategy without legacy constraints.
• Access to a unique, growing dataset from real-world sorting facilities.
• Utilizing the latest frameworks and a state-of-the-art tech stack (PyTorch 2.x, FSDP, Hydra, W&B, DVC, timm).
• Substantial compute resources for pretraining runs.
• Close collaboration with research partners (e.g., THWS Würzburg) and the opportunity to publish research findings.
• Working within a dynamic, interdisciplinary startup team with significant responsibility and creative freedom.
• Short decision-making paths and direct communication.
• Developing technology that contributes to AI, recycling, and the circular economy.