LLM Inferencing As-a-Service

• SNET Compute

Development and release of LLM inferencing as-a-service accessible via ASI Cloud platform (formerly CUDOS Intercloud) and onboarding first customers. The service is jointly developed with CUDO and will set the foundation for SNET Compute AI Services offering. The service will utilise the components listed above including LLM hosting, KV Global Cache (LLM optimisation) and in the future will include End-to-end Gen AI Environment encryption (encrypted inferencing), serverless  as well as speech and image processing modules.

Serverless Platform MVP

• SNET Compute

Clarifying business requirements and the possibility of reusing the serverless SingularityNET-orchestrator POC. Optimizing and improving the current implementation. Adding DevExp for tools in the current implementation. Defining functional requirements for transforming it into a serverless platform.

Banxa-based payment system

• Platform

• Fiat and Crypto payments flow for paying for Platform AI Services via Banxa provider, Banxa - MPE Controllers, etc.
• Fiat and Crypto payments flow for paying for HaaS on the platform via Banxa provider.

LLM Hosting

• SNET Compute

Creating a fault-tolerant hosting system for large language models. Scaling mechanisms, token counting, integration of Cudo authorization, and payment tools with Cudo. Adding metrics, alerts, and fault-tolerant mechanisms for model hosting. Using distributed backends for models similar to RAY.

AI Synthesizer

• Platform

Audio and metadata preprocessing. Dataset preparation. Fine tuning and training of the MusicGen model. Experiments with model training. Development of model and service testing logic. Service packaging and support.

ASI: Create Release 1

• ASI

Create provides a versatile and scalable platform where developers can launch innovative AI projects with ease. It combines a user-friendly interface with powerful backend tools, including the ASI-1 LLM, Qwello Researcher and other neural net tools, alongside knowledge graphs and OpenCog Hyperon based reasoning, to facilitate rapid development and deployment. The platform is designed to handle the full lifecycle of AI projects—from initial concept to full-scale production—and supports decentralized capital raising and data trading. This ensures that every project is built on a solid foundation and integrated into a broader decentralized ecosystem. Looking forward, Create is set to undergo significant enhancements with the integration of more advanced Hyperon-based proto-AGI features later this year.

Publisher Portal v3

• Platform

This task includes the automation of the transfer of the UI interface developed within the UI Sandbox directly into the general flow of onboarding and editing the service (now we have to ask to download .zip file and upload separately in Publisher Portal), as well as the potential (not approved) integration of the Publisher Portal into the Marketplace, as a prototype of the personal account of organizations. The key task is to unite the Platform tools within the general flow of registration and editing of the service, including HaaS, UI Sandbox and its improvements.

Add training functionality v2 to Marketplace

• Platform

Updating the training functionality will simplify and speed up the process of adding new models for users. Now it will be possible to get requirements for datasets, and interaction with models will be more convenient and require fewer signatures from the user. It will also be possible to make paid calls to the training functions. The system will no longer need an additional request to get the current state of the service.

Release of PRIMUS components

• OpenCog Hyperon

including PLN, ECAN, MOSES and Pattern Miner