Roadmap
Here we provide an overview of current and future focus areas for further development of the functionality of SciLifeLab Serve. The team behind the platform regularly reviews and updates the long-term plan. When planning work we always take into account the feedback that we receive from our users through various channels when planning our work. Feel free to get in touch with us with questions, comments, and suggestions.
Current work
Metadata and FAIR improvements
We will be collecting additional information about applications and models published on SciLifeLab Serve (such as list of authors and source of funding). We will also be registering Digital Object Identifiers (DOIs) to allow tracking of metadata and citation of applications and models.
Improved workflow for app creation & better feedback
For newly created applications, a multi-step process will be launched that will provide feedback if there is an issue with the provided Docker image or elsewhere. This way we hope to make catching issues and troubleshooting easier for app authors.
Access to an API endpoint to interact with LLM(s)
We will provide an API endpoint for application developers that will allow their apps to interact with one or more LLMs. The apps will then need to be made available through SciLifeLab Serve.
Near-term work
Cluster internal security improvements
We will focus on improving security on the cluster where SciLifeLab Serve is hosted. For example, already now applications and models in projects are isolated from those in other projects; we will make this separation stronger.
Login using university credentials
We will allow users to log in to SciLifeLab Serve using their Swedish university credentials (and without the need to register separately). Those who do not have these credentials will be able to log in using a login and a password as before.
Privileged user category
We will implement a new category of users that will have more permissions than regular users. For example, the privileged users will be allowed to create a higher number of projects and apps as well as to get more storage per project.
Long-term work
Improved model-serving functionality
We will improve the user experience and useability of deploying and using machine learning models through SciLifeLab Serve.
LLM-empowered application creation
We will integrate an interface allowing to build data science applications by interacting with an LLM through a chat interface. These applications can then be published on SciLifeLab Serve.
Past work
Improved web accessibility of the website
We will review and update our webpages to meet the Web Content Accessibility Guidelines (WCAG) 2.2.
Support for model deployment with GPUs
We will allow users to publish models with access to GPUs to run inferences. We will start with a few pilot projects before making it widely available. Get in touch with us if you are interested in being a pilot user.
Integration of MLFLow for ML experiment tracking
We will allow users of SciLifeLab Serve to launch their own instances of the platform MLflow (opens in a new tab) that allows to keep track of models and artifacts in machine learning work. Users will be able to send data to their instance from anywhere where they are doing training or analyses.
Page last updated:
