Project Overview
Developed a vision- and NLP-based validation reference project for neuromorphic processor (NPU) clusters.
Key Responsibilities
- Set up infrastructure with Kubernetes, MinIO, Harbor, DVC, Docker, and Kubeflow.
- Designed AI project architecture and implemented source code using Cookiecutter.
- Developed vision dataset preprocessing modules.
- Automated hyperparameter tuning with Kubeflow Katib.
- Built Kubeflow Pipelines for data preprocessing, augmentation, training, and deployment.
- Deployed and versioned models using BentoML.
Achievements
- Independently completed the project within a short timeline, contributing to company revenue.
- Enhanced reusability and scalability by containerizing components and organizing YAML scripts.