NavLife

Navigate your life in a new city

Comment

Inspiration

As international students, we often face a variety of challenges while adjusting to a new environment — from finding safe travel routes and affordable food to navigating public transport or simply figuring out how to dress appropriately for the weather. These seemingly small everyday questions often go unanswered, especially when you're new and traveling alone. That’s why we built NavLife — a one-stop solution to help students get critical, contextual information easily and reliably.

What it does

NavLife offers four core features:

  1. Safest Route Planner When traveling at night, safety is critical. NavLife uses real crime data to identify and recommend the safest routes. Zones are visually marked as red (high risk), yellow (moderate), or green (safe), helping you make informed decisions about where to walk or drive.

  2. Fastest & Cheapest Route Finder Simply ask how to reach your destination, and our AI agent will find the fastest and cheapest transit option using real-time schedule data (bus, BART, etc.). It plans intelligently — so you don’t have to.

  3. Smart Food Discovery Craving something tasty without breaking the bank? NavLife recommends nearby, affordable food joints based on your preferences (e.g., “I wanna have something sweet”) and real-time restaurant data.

  4. Weather & Outfit Assistant Don’t get caught off guard by unexpected weather! NavLife tells you the current temperature and suggests what to wear (e.g., “Wear a hoodie — it will be windy later tonight”) to keep you comfortable and prepared.

How we built it

NavLife is made up of four independent features, each powered by dedicated tools and frameworks:

  1. Food Discovery We used Gemini-Pro (Google) as the core language model within the Letta framework to interpret natural language food queries. The Google Places API was integrated as a tool, allowing the agent to fetch nearby restaurants based on filters like cuisine, budget, and open hours. This combination enabled personalized, real-time food recommendations.

  2. Weather Awareness The weather feature was built using Letta and powered by Gemini-Pro (Google) to format responses. The agent used the OpenWeatherMap API to fetch current temperature, humidity, and conditions. The weather data is displayed in the interface but does not yet influence other features.

  3. Safe Route Planning This feature also uses Letta with Gemini-Pro (Google) to analyze police incident data from the San Francisco Police Department. A custom tool scores different geographic zones by safety level. The agent uses this data to suggest walking or biking routes that avoid high-risk areas.

  4. Transit Intelligence We used CrewAI to create a set of agents—each powered by Gemini-Pro (Google)—that collaborate to handle public transit tasks. These agents load transit datasets, locate nearby stops, assess routes, and recommend optimized travel plans. The crew runs in sequence to simulate multi-step reasoning.

Challenges we ran into

Integrating Gemini with both Letta and CrewAI required strict schema definitions and careful error handling. Route safety logic had to be built from scratch using incident data and spatial scoring. Finally, merging work from four different branches required planning and modular design to avoid breaking the system.

Accomplishments that we're proud of

We delivered four working, real-time features under one interface. Each feature is powered by agents, runs live, and connects to real-world data. We migrated to Gemini, integrated APIs from Google and OpenWeather, and turned crime datasets into practical route recommendations. Most of all, we built something that international students could use on day one in a new city.

What we learned

Agent frameworks are powerful but demand precision. Letta made it easier to build focused, stateful agents. CrewAI helped us split complex tasks across multiple specialized agents. Real-time APIs add depth but require fallback logic and strict formatting. Team modularity helped us move fast—each person could build independently, and it all came together cleanly.

What's next for NavLife

We plan to integrate weather into food and transit logic, so users get smarter suggestions in bad conditions. We'll add voice interaction using LMNT or Vapi. We'll persist user preferences across sessions to offer more personalized advice. And we aim to launch this for students across other cities, starting with UC campuses.

Built With

Share this project:

Updates