.png&w=256&q=75)
🤝 Top Collaborators
.png&w=640&q=75)
.png&w=640&q=75)
🤓 Latest Submissions
On-Prem AI Unified Namespace for OT Automation
Industrial telemetry lives in OPC UA servers, while crucial context is buried in unstructured documentation. Our hackathon prototype fuses the two entirely on-prem. A Python Asset Hierarchy Extractor runs Llama-3-8B on Groq silicon, scrapes every Wiki.js page, deduplicates devices, browses the OPC UA address space and links tags to assets. The resulting YAML map feeds a Python UNS Configurator, which generates Node-RED flows and publishes Sparkplug B topics via a local Mosquitto broker—instantly creating a self-describing Unified Namespace. In parallel, a RAG Importer writes the sanitised context into OpenWebUI’s vector database so LLM answers stay grounded in plant knowledge without exposing control-system secrets. A .NET Historian subscribes to Sparkplug B, writes raw rows and aggregates to Snowflake, and caches metadata in Redis. Two lightweight industrial APIs make the data actionable: a Sparkplug-B MCP (Python, paho-MQTT) for DCMD commands, and a Historian MCP (Python, Snowflake SDK) for reads. Operators use OpenWebUI with Llama-3-70B to chat, pull trends or send set-points—completely offline. For repeatable actions they type plain-language rules; a Business Rule Agent (compound-beta) converts them to Python, and a Business Rule Runtime stored in Snowflake executes them every few seconds through the same MCP layer. Everything runs in Docker on Vultr Compute yet remains air-gapped: inference, vector search, MQTT traffic and historian storage never leave the plant network. The result is contextualised data, natural-language insights and safe no-code automation, delivered by a small AG-Solution MES/MOM team working after hours.
8 Jul 2025
AI-Powered Smart School Connectivity
Access to high-speed internet is essential for modern education, yet many schools, especially in remote areas, lack proper connectivity. Traditional approaches, such as building new 5G towers, are expensive and inefficient. Our project leverages Artificial Intelligence and mathematical modeling algorithms to classify schools based on their connectivity needs and generate the most cost-effective scalable expansion strategy. Using AI-powered clustering techniques, we analyze school locations, existing network coverage, and infrastructure constraints to categorize schools into four main types: 1 - Fully Connected Clusters: Single or groups of schools where all have internet connection. 2 - Partially Connected Clusters: Groups of schools where at least one already has internet access. 3 - Isolated Clusters: Multiple schools with no existing internet infrastructure. 4 - Remote Standalone Schools: Single schools lacking any connection. Our mathematical optimization model then determines the optimal network expansion plan, recommending the most efficient infrastructure improvements. By intelligently leveraging existing connectivity and optimizing the placement of additional access points, we provide universal school connectivity with minimal cost and maximum efficiency. Key AI Features and Benefits - AI-Driven Classification & Clustering: Algorithms analyze geographical, infrastructure, and network data to classify schools and group them strategically. - Network Optimization: Mathematical models forecast the best expansion strategy, ensuring connectivity with the least infrastructure investment. - Cost-Efficient & Scalable Solution: Data-based decision-making enables rapid scalability and adaptation to different regions. This project is designed for governments, NGOs, and telecom providers seeking AI-driven, data-backed solutions to bridge the digital divide in education efficiently and affordably.
2 Mar 2025
