Hugo Bowne-Anderson - Full-stack Machine Learning and Generative AI for Data Scientists

Discover how to deploy machine learning models at scale with Full Stack ML and Generative AI. Learn about reusable workflows, scalable applications, and collaboration with data engineers for seamless prototype-to-production pipelines.

Key takeaways

Full stack machine learning involves moving between prototype and production, and requires versioning data and models.
Metaflow is used to create reusable workflows that can be easily adapted to multiple use cases.
A combination of Python, Argo, and Kubernetes can be used to build scalable machine learning applications.
The speaker emphasizes the importance of keeping the data engineering layer separate from the data science layer.
Generative AI models can be used to create products quickly, and Metaflow can help in building these products.
The speaker encourages experimentation and prototyping, and highlights the need to consider versioning and scalability when building machine learning applications.
Metaflow and Airflow have always allowed for bespoke integrations, and the speaker emphasizes the importance of interoperability.
The speaker also mentions the need for infrastructure, including data engineering, models, and code.
He also points out that data scientists and engineers need to work together to deploy applications.
A cloud workstation can be used to build machine learning applications, and Metaflow can help in this process.
Forecasting can be achieved by using predictive models, and the speaker shows an example of this in his presentation.
Embeddings can be used to convert text data into high-dimensional vector spaces, which can be used for prediction.
Animations can be used to visualize the output of a machine learning model.

Hugo Bowne-Anderson - Full-stack Machine Learning and Generative AI for Data Scientists

More talks