The best of both worlds: Combining Python and Kotlin for Machine Learning

Bridge the gap between data science and engineering. Use Python for ML experimentation and Kotlin for robust, production-ready APIs.

#1about 5 minutes

The production gap in machine learning

Most machine learning models fail to reach production due to the disconnect between data scientists and software engineers, and the complex MLOps lifecycle required.

#2about 8 minutes

Data exploration and analysis with Kotlin in Jupyter

Use the Kotlin kernel in Jupyter notebooks with libraries like DataFrame and Let's Plot to perform type-safe data analysis and visualization.

#3about 3 minutes

Building neural networks with the Kotlin DL library

Define and train a neural network model using the Kotlin DL library, but be aware of current limitations like incompatibility with ARM-based Macs.

#4about 4 minutes

Deploying ML models as a web service with Ktor

Serve a pre-trained ONNX machine learning model with a lightweight web service using the Ktor framework for easy integration into production systems.

#5about 3 minutes

Choosing between Python and Kotlin for ML tasks

Use Python for its mature ecosystem in model development and experimentation, while leveraging Kotlin's type safety and performance for data pipelines and model serving.

#6about 2 minutes

Q&A on Kotlin for machine learning

The speaker answers audience questions about Kotlin DataFrame internals, integration with other frameworks, and the connection between the Kotlin and Python ecosystems.