[Machine Learning Engineer] Edge AI / Model Optimization (Kyoto)English summary from Japanese

Must-haves

• Understand and modify PyTorch image analysis models for edge device compatibility.
• Perform model conversion for INT8 quantization, ONNX export, and inference runtimes.
• Identify and replace unsupported PyTorch operations with alternatives for edge deployment.
• Evaluate model performance (accuracy, speed, memory usage) on edge devices.
• Implement Python scripts for evaluating image analysis models using metrics like Precision, Recall, IoU, mAP.
• Build and manage reproducible evaluation environments.
• Utilize Git/GitHub for code management, reviews, and team development.
• Document and share findings on model conversion, quantization, and evaluation results.

Nice-to-haves

• Experience with model optimization and acceleration for edge devices.
• Expertise in quantization, ONNX conversion, and handling unsupported operations.
• Experience in MLOps, including quantitative evaluation of trade-offs (accuracy, speed, memory) and ensuring evaluation reproducibility.
• Ability to investigate and validate new conversion tools, inference runtimes, and edge AI technologies.
• Strong communication skills for sharing progress, risks, and improvement plans.

Tech stack

Work style

Onsite in Kyoto