ML Deployment
Description: This lecture will start looking into the data dimension of the ML concept and will emphasise on the importance of data-orientation. We first define the concept of data, the associated challenges, and provide examples of data collection processes. We then define a data science methodology to iteratively build the datasets that will feed our machine learning models. This lecture explores the first step of this methodology, "data access".Department: 34th International Symposium on Statistics
Institution: Universidad de Nariño
Date: July 30, 2025
Hours: 1.5
From: 8:00 am
To: 09:30 am
Resources
Books
- Kleppmann, M. (2017). Designing Data-Intensive Applications. O’Reilly Media
- Huyen, C. (2022). Designing Machine Learning Systems. O’Reilly Media
Papers and Reports
- Sculley, D., et al. (2015). Hidden technical debt in machine learning systems
- Breck, E., et al. (2016). The ML test score: A rubric for ML production readiness and technical debt reduction
- Kabi J., Maina C. (2021). Leveraging IoT and Machine Learning for Improved Monitoring of Water Resources - A Case Study of the Upper Ewaso Nyiro River
- Zittrain J. (2022). Intellectual Debt: With Great Power Comes Great Ignorance
- Paleyes, A., et al. (2020). Challenges in deploying machine learning: a survey of case studies
- Cabrera, C., et al. (2022). MAACO: A Dynamic Service Placement Model for Smart Cities
- Cabrera, C., et al. (2023). Machine Learning Systems: A survey from a Data-Oriented Perspective
Web
- The Data Dichotomy: Rethinking the way we Treat Data and Services
- Deploying Machine Learning using Flask
- MLOps: Continuous delivery and automation pipelines in machine learning
- Kubeflow: Machine Learning Toolkit for Kubernetes
- MLflow: An open source platform for the machine learning lifecycle
- TensorFlow Serving
- TorchServe: Model serving for PyTorch
- Seldon Core: Cloud native machine learning deployment
- Weights & Biases: MLOps platform
- Neptune.ai: Experiment tracking and model registry