Senior Computer Vision Developer

Summary

Join a dynamic team at Miovision as a Senior Computer Vision Scientist and contribute to advancing the field of computer vision through innovative approaches to simulated environments, 3D scene understanding, and robust model development.

Requirements

• 7+ years of experience developing computer vision and machine learning, PhD preferred
• Deep expertise in 3D computer vision , scene understanding, and geometric deep learning
• Experience with photorealistic rendering, physics-based simulation, automatic ground truth generation for instance and semantic segmentation, as well as camera homography, optimization, and image processing
• Strong software skills, particularly with Python and C++ on Linux
• Extensive experience with simulation environments (CARLA, AirSim, Unity, Unreal Engine) , their integration with real-world data pipelines, and automated generation of precise ground truth segmentation maps
• Strong background in bridging simulated and real-world domains through advanced adaptation techniques
• Experience in developing and maintaining robust data quality assessment frameworks and expertise in designing and implementing validation methodologies for computer vision systems
• Proven track record of translating research papers into practical, production-ready solutions

Responsibilities

• Lead the development of advanced simulation environments using platforms like CARLA and Unreal Engine, ensuring high-fidelity representation of real-world scenarios
• Design and implement automated pipelines for generating high-quality ground truth segmentation maps from simulated environments, ensuring precise pixel-level annotations for training and validation
• Guide technical decisions around data annotation strategies and quality control processes
• Design and implement comprehensive data quality frameworks that span both simulated and real-world datasets, including automated validation systems and data integrity checks
• Develop and optimize 3D scene understanding capabilities and multi-view geometry, including depth estimation, object localization, spatial relationship modelling, and automated generation of precise segmentation ground truth from synthetic data
• Create robust validation methodologies to ensure simulation-trained models perform reliably in real-world conditions
• Bridge the gap between simulated and real-world data through advanced domain adaptation techniques and careful calibration processes
• Research and implement novel approaches for translating academic breakthroughs into production-ready computer vision solutions
• Design and maintain data augmentation and synthetic data generation pipelines that realistically capture edge cases, environmental variations, and produce accurate ground truth segmentation maps
• Establish quality metrics and validation procedures for assessing model performance across simulated and real-world environments
• Lead initiatives to transform research concepts into scalable, production-ready solutions
• Participate in technical roadmap discussions and estimation exercises

Preferred Qualifications

• Experience with modern data annotation platforms (e.g., V7, Scale AI, Labelbox), designing efficient labeling workflows, and implementing automated ground truth generation pipelines, including creating annotation guidelines, quality assurance processes, and managing annotation team workflows
• Experience with the machine learning operations (MLOps) workflow and in developing robust evaluation metrics for computer vision systems
• Track record of publishing in top-tier computer vision conferences (CVPR, ICCV, ECCV) and a desire to contribute to writing technical reports and/or research papers
• Experience with edge-computing such as Jetson Xavier, TX2, or FPGA, and video pipeline performance optimization
• Familiarity with running and training inference engines using at least one of TensorFlow, OpenCV, PyTorch, TensorRT, ONNX