It’s a wrap — Final delivery at ACDC 2026 🚀

Posted on by

This is our final contribution for the ACDC2026 Hackathon, our final delivery.
We made a video of the end to end solution, enjoy the video.

Image bellow describes the overall design of the solution.

🧱 Redstone Realm

Showing Jetson Nano Edge AI LLM providing guidance for the customer off-grid.
  • Built a real, working solution while actively exploring new Microsoft platform capabilities
  • Used in-box AI such as Prompt Columns and Copilot to embed AI directly into the data model and user experience
  • Grounded AI output in structured data to keep interactions predictable and explainable
  • Used Code Apps to experiment with new ways of building user-friendly experiences and validate ideas quickly
  • Experimented with Edge AI using an NVIDIA Jetson Nano to run LLMs closer to execution
  • Explored trade-offs between edge-based and cloud-based AI through hands-on experimentation

Redstone Realm, for us, was about building, testing, learning — and pushing understanding forward using real tools on real platforms.

Relevant blogpost:
Existential Risk: intelligence without agency, Nvidia Jetson Nano, Glossy Pixels | Arctic Cloud Developer Challenge Submissions

🛡️ Governance & Best Practices

  • Stored all secrets in Azure Key Vault, accessed at runtime by Azure Functions and Power Automate via environment variables
  • Used a clear DEV / TEST / PROD environment strategy with a structured ALM setup for predictable deployments
  • Maintained clear architectural separation between UI, integration, and execution, with deterministic and testable backend logic
  • Applied consistent naming conventions across fields, flows, and assets
  • Used a medallion data structure (raw, refined, curated) to ensure data quality and traceability
  • Used Copilot as an assistive, explanatory layer — not an autonomous decision-maker

Governance was built in from the start to ensure the solution is secure, maintainable, and trustworthy beyond the hackathon.hackathon.

Relevant blogposts: ALM implemented

🧠 Data, AI & Analytics

Even in a hackathon setting, we designed with structure and responsibility in mind.

From raw blocks to blazing insights: use Microsoft Fabric to take messy data through a structured refinement process, model it into trusted semantic layers, unlock visual storytelling with Power BI, and build a foundation with Fabric IQ that helps both AI agents and data scientists uncover the real value in your datasets. If something doesn’t add value, keep polishing until it sparkles! 💎

Relevant blogposts: From raw blocks to data diamonds

⚡ Low-Code

  • Used low-code to move fast while keeping structure and maintainability intact
  • Built a back-office Model-Driven App for governance, search, and operational overview
  • Used Prompt Columns to embed AI directly in the data model and enable predictable Copilot behavior
  • Leveraged new Power Platform capabilities to deliver advanced functionality quickly and securely
  • Established an analytics foundation using Microsoft Fabric with a medallion architecture (raw, refined, curated)

    Relevant blogposts: OneFlow and LINK Mobility Sponsor Badge and more, Go With The Flow | Arctic Cloud Developer Challenge Submissions


🧑‍💻 Code Connoisseur

  • Built a Code App using vibe coding, outside traditional Model-Driven and Canvas patterns
  • Implemented Azure Functions in C# (.NET 8) with RCON API integration to a Minecraft server
  • Ensured backend logic is deterministic, testable, and decoupled from UI and AI
  • Experimented with Edge AI using NVIDIA Jetson Nano, leveraging Linux shell tooling and low-level configuration
  • Explored trade-offs between edge-based LLMs and cloud-hosted AI services
  • Kept business logic in code, with clear separation between experience, AI, and execution

Code was used deliberately — where control and predictability mattered most.

Relevant blogposts: OneFlow and LINK Mobility Sponsor Badge and more

    🌍 Digital Transformation

    • Built a solution that starts from user intent rather than technical specifications
    • Transformed intent into structured data that can be reasoned about, adjusted, and reused
    • Used AI and Copilot to explain consequences and trade-offs instead of automating decisions
    • Connected business logic, data, and visualization into a continuous feedback loop
    • Used Minecraft as a visualization engine to make outcomes tangible and easy to understand
    • Demonstrated how low-code, pro-code, and AI can work together to support better decisions

    The transformation was not just technical — it changed how users understand and act on complex decisions.

    Relevant blogposts: OneFlow and LINK Mobility Sponsor Badge and more, NASTY! If it doesn’t work, expose it to the world | Arctic Cloud Developer Challenge Submissions