NemoCode.io

Look of the webapp

Inspiration

During my recent trip to Arizona, I had to figure everything out on my own. Driving? or Flying? Should I carry warm clothes or cold? What are the best places to visit in 4 days? It was a hassle. Inspired our team to build NemoCode.io - Nemotron Itinerary Agent, an autonomous AI travel planner that goes far beyond a normal chatbot. A chatbot is reactive, you need to input and command it every single time. Our Nemo is a proactive fishy - it reasons through budgets, weather, routes, and availability to build a full, personalized itinerary in real time for you.

What it does

Nemo thinks like a human travel planner but moves with AI speed. Using NVIDIA Nemotron, we created an agent that can parse a simple prompt like

“Plan a 3-day Austin trip for two people next weekend under $800 with live music and BBQ” and autonomously plan the entire journey for us tailored to your experience.

How we built it

We built a full-stack intelligent system with: Frontend: Tailwind + JavaScript for clean, responsive UI Backend: Flask REST API for managing user inputs and tool orchestration AI Core: NVIDIA Nemotron Agent that performs three key roles: -> Planner – interprets the user’s request and creates an execution graph

-> Executor – calls real-world APIs (maps, weather, hotels, music)

-> Synthesizer – merges all results into a personalized, budget-aware itinerary APIs integrated: Google Maps, Places, Weather, and Gemini for city history Visualization: Live itinerary map + tabs for schedule, budget, clothing, music, and culture We designed a modular agent architecture where Nemotron reasons at every stage — checking weather, optimizing cost, and even generating style and music recommendations to match the trip vibe.

Challenges we ran into

Getting Nemotron to consistently output structured JSON for planning steps. Keeping budgets and constraints coherent through multiple tool calls. Integrating multiple APIs seamlessly under one orchestrated agent. Designing a UI that visually explains the agent’s reasoning process in real time.

Accomplishments that we're proud of

The final demo shows our Nemo autonomously planning a trip, re-planning when weather changes, and logging every reasoning step. Judges can literally watch it “think.”

What we learned

How to make a large model act, not just chat — giving Nemotron a “planner–executor–synthesizer” workflow. The power of multi-step reasoning and structured JSON prompting for reliable API orchestration. How to bridge design and intelligence: combining AI output with human-friendly UX. Building fallback mocks so the entire system can run offline for demos.

What's next for NemoCode.io

Nemo can be much more than a basic travel agent. It can integrate experiences, personalize your recommendations according to your taste - a cosy trip or an adventurous hike? Nemo knows what you want and will ensure you get the best deal out there. Nemo can be your travel companion and help you in real-time with anything you want to create and plan during your trip.