Sponsor Presentation - Johannes Messner: Modern, typed Python for (multimodal) ML

Python

Discover how modern, typed Python enables machine learning with multimodal data using advancements in type hinting, Pydantic, FastAPI, and more.

Key takeaways

Modern, typed Python for machine learning is possible with advancements in type hinting, making it easier to work with multimodal data.
Pydantic and FastAPI allow for type-safe APIs and automatically coerce types, reducing the risk of errors.
The Python ecosystem is rapidly evolving to support type hinting, with tools like MyPy and Typer, making it easier to write type-safe code.
Vector databases like Qdrant, Elasticsearch, and HNSWlib are well-suited for storing and querying vector data, and can be integrated with Python machine learning models.
The duck test principle can be applied to ensure type-safety, where a type is not just defined by its shape, but by its behavior.
Docker 8 supports generics and can be used to build vector databases, making it a valuable tool for machine learning applications.
When working with multimodal data, it is often necessary to use multiple libraries and frameworks, making it important to have a clear understanding of the relationships between them.
Typing can help reduce errors and make code more maintainable, especially in machine learning applications where data types can be complex and nuanced.

More talks