Olympus

Olympus is an intelligent infrastructure orchestration platform that unifies Terraform automation with AI-driven decisioning through an NVIDIA-routed Model Context Protocol (MCP) server.

Comment

Inspiration

Cloud infrastructure is powerful but difficult to express and manage due to the constant monitoring and endless logs. We wanted to build something that lets people describe what they mean and have the system take care of the technical depth. With NVIDIA and NorthMark expanding cloud automation, Olympus aims to make cloud feel intuitive.

What it does

Olympus turns natural-language requests into Terraform deployments using a Model Context Protocol (MCP) pipeline. It intelligently interprets intent and different functionalities, and generates infrastructure plans while executing them safely. You bring your environment, Olympus manages it.

How we built it

We built a robust frontend responsible for showing relevant data flows and allowing users to inquire about the resources via NVIDIA's reasoning nemotron-nano-9b-v2 model. The LLM enhances interpretation and task planning, communicating with a Flask backend that relays information with an in-house and from-scratch MCP server responsible for Terraform orchestration. The terraform server is dockerized to maintain consistency and repeatability. The response from the request is returned via frontend visual output and resources are managed and reconfigured.

Challenges we ran into

Translating ambiguous language into precise Terraform syntax was tedious. Coordinating multiple services to work reliably and synchronously under time pressure required careful planning and implementation. Additionally, we had to ensure safe deployment and rollback safeguards to avoid cloud chaos and excess overhead.

Accomplishments that we're proud of

We successfully built an end-to-end natural-language-to-infrastructure workflow. We created a modular system ready to scale with new models and cloud environments. Most importantly, we made the more undecipherable and hard-to-interpret cloud provisioning tasks feel more interactive and human.

What we learned

At the core of large-scale infrastructure lies simplicity. We have learned that translating user goals requires a flexible approach. This holds true across all levels, from foundational design principles like decomposition to the more nuanced layers of orchestration.

What's next for Olympus

We plan to expand to multi-cloud support and add collaborative deployment planning. Future versions will deepen NVIDIA-powered reasoning for more consistent architecture suggestions. Excited for Olympus to evolve!

Built With

Share this project:

Updates