K2 Platform | HackUDC 2026

K2 Platform is an advanced data orchestration and intelligence hub developed for HackUDC 2026.

Comment

Inspiration:

Modern organizations are drowning in data but starving for actionable insights. We observed a recurring bottleneck: developers and analysts spend 80% of their time wrestling with siloed databases and "cleaning" fragile connections just to get a glimpse of the big picture. We wanted to build a "Single Source of Truth" that wasn't just a static database, but an intelligent hub. Inspired by the rugged efficiency of the K2 mountain, we built K2 Platform to be the peak of data orchestration.

What it does:

K2 Platform is a comprehensive data intelligence hub. It uses Data Virtualization to connect to multiple disparate sources (SQL, NoSQL, APIs) without moving the data. On top of this layer, we integrated a Generative AI interface that allows users to query complex organizational data using natural language. No more complex SQL joins; just ask the K2 Chatbot, and it fetches the real-time data through Denodo serving it through a high-performance React dashboard.

How we built it:

The project is built on a robust, modern stack:

- Data Layer: Denodo VDP acts as the virtualization engine, unifying our data sources into a single logical layer.
- Intelligence: We utilized the Denodo AI-SDK to implement RAG (Retrieval-Augmented Generation), allowing our LLM to have context-aware conversations with the virtualized data.
- Backend: A FastAPI (Python 3.12) server manages the business logic and serves as the bridge between the AI and the frontend.
- Frontend: React 19 with Vite and Tailwind CSS provides a sleek, responsive user interface.
- Infrastructure: We used Nix to create a fully reproducible development environment, ensuring that "it works on my machine" applied to the whole team instantly.

Challenges we ran into:

The biggest hurdle was the initial configuration of the Data Virtualization layer mapping disparate schemas into a unified view that the AI could understand without hallucinations. We also faced a race against time to integrate the Denodo AI-SDK with our FastAPI streams to ensure the chatbot felt "snappy" and real-time rather than lagging during data retrieval.

Accomplishments that we're proud of:

- The "Nix" Factor: We managed to get a 4-person team up and running with a complex environment (Python, Node, Denodo connectors) in minutes using a single nix-shell command.
- Zero-Copy Integration: We successfully queried different data types through Denodo without writing a single ETL (Extract, Transform, Load) script.
- The UI/UX: Achieving a professional, "SaaS-ready" look and feel within the 48-hour limit of HackUDC.

What we learned:

We delved deep into the world of Data Virtualization, learning that moving data is often the worst way to handle it. We also leveled up our skills in RAG architectures, specifically how to structure metadata so that LLMs can generate accurate queries. Most importantly, we learned the value of a declarative environment, Nix saved us hours of debugging "environment drift."

What's next for K2 Platform:

- Predictive Analytics: Moving from "what happened" to "what will happen" by integrating Machine Learning models directly into the Denodo layer.
- Edge Integration: Extending the platform to handle IoT data streams in real-time.
- Granular Permissions: Implementing OIDC and Row-Level Security (RLS) within the Denodo layer to make K2 enterprise-ready for sensitive data.

Built With

  • denodo-ai-sdk-(rag-&-chatbot
  • denodo-vdp-(virtual-data-port)
  • docker
  • fastapi-(python-3.12)
  • grafana
  • javascript
  • nix
  • node.js
  • react-19-(vite)
  • sqlite
  • tailwind-css
  • vql
Share this project:

Updates