Technology Fridays: Algorithmia CODEX Wants to be the Runtime for Enterprise AI
Welcome to Technology Fridays! Today we are going back to deep learning land with one of the most exciting new releases in the space: Algorithmia CODEX. recently, I’ve published a couple of articles that highlights Algorithmia Algorithm as a Service(AgaaS) model as one of the most creative offerings in the artificial intelligence(AI) market. CODEX is the newest addition to the Algorithmia platform and one that expands its capabilities into the enterprise space.
CODEX is trying to address one of the biggest challenges in AI solutions by providing a runtime that enables the execution and management of AI applications. From that perspective, CODEX provides a cloud-agnostic runtime that streamlines the lifecycle management of AI models implemented using diverse programming languages such as Java, Scala, Ruby, NodeJS, Python and others. By cloud-agnostic, we are referring to CODEX’s capability to operates on different cloud platforms such as AWS, Azure or Google Cloud we well as on-premise environments powered by OpenStack. CODEX also actively leverages as part of its infrastructure.
One of the main capabilities of Algorithmia CODEX is its ability to generate REST APIs directly from AI models. CODEX APIs receive request from client applications, execute the corresponding algorithm and return the results in the from of JSON payloads. CODEX also provides client libraries that abstract the integration with the generated APIs.
The core CODEX infrastructure is composed of three types of nodes. API nodes are responsible for hosting the APIs generated from registered algorithms. APIs nodes receive request from client applications and pass them to worker nodes. Algorithms deployed to the CODEX platform are hosted in worker nodes. That type of nodes are, essentially, containers clusters in which each container hosts a specific algorithms. CODEX includes two main types of worker nodes: GPU and CPU depending on the underlying hardware infrastructure. The third type of node supported by CODEX are Web Nodes that are responsible fro hosting the platform’s web portals such as its Admin Panel, the Legit git hosting service or Pyrometer metering service.
CODEX extends the lifecycle of AI applications beyond just algorithms. The platform also enables the registration and management of datasets from heterogeneous sources such as HDFS or SQL-based databases. That capability allows the management of AI models together with its training and test data sources.
Monitoring and instrumentation are other relevant capabilities of the Algorithmia CODEX platform. From APIs, to AI models to data sources, CODEX provides an end-to-end visibility into the execution of AI applications. Similarly, CODEX enables robust access control and data privacy capabilities across all tiers of the platform.
Algorithmia CODE is operating in the emerging market of hybrid AI runtimes. Innovative startups such as Bonsai, H2O.ai or Bitfusion are actively competing for developers’ mindshare in that market. Native cloud machine learning services such as AWS ML, Google Cloud ML or Azure ML can also be considered CODEX’ competitors in cloud environments.