Computer Vision Engineer

Role overview Build and improve the computer vision data-generation and model-evaluation workflows used for security, defense and robotics applications. You’ll take work through the full cycle—from generating and annotating examples to training models, measuring performance and refining the pipeline—in a small team with direct input into what ships.

Responsibilities - Create and maintain ComfyUI workflows for generating training data across relevant scenarios. - Fine-tune LoRA models on domain data such as equipment, vehicles and environments. - Train and benchmark YOLO detectors and vision-language or vision transformer models using synthetic and real data. - Compare model performance on generated and real-world data, then use findings to improve generation quality and dataset coverage. - Build fine-tuning workflows for vision-language tasks such as captioning, grounding and classification. - Maintain annotation accuracy and dataset quality as work is released.

Requirements - Strong Python and PyTorch skills. - Experience training, fine-tuning and deploying YOLO models, version 8 or later. - Familiarity with vision transformers and vision-language models. - Ability to build and modify ComfyUI workflows. - Experience fine-tuning LoRAs with SDXL, Flux or comparable models. - Working knowledge of YOLO and COCO annotation formats.

Nice to have - Knowledge of diffusion model components or tools such as ControlNet. - Experience in CCTV, defense, robotics or synthetic data. - Deployment experience with CUDA, Triton, ONNX or TensorRT. - Open-source contributions and comfort delivering work in a small team.

Benefits and work setup Remote role, with EU-friendly working hours. The position offers GPU access for training and experimentation, ownership of technical direction, and direct feedback from customers.