WayMore

Type a start and a destination or tap the demo chip for Brickell → Wynwood.
Touring counts the places passed, shows the line being spoken right now, and gives you a pause button and separate mute chips for landmarks
Compare. Fastest vs. Scenic with the honest cost of the detour. Switching redraws the map from data already in hand
Browse views. Tap a pin or a card for the photo, the rating and review count and the description: cuisine and price level for restaurants

Inspiration

The idea for WayMore came from a friend of mine who is always driving down from Gainesville. On one of his trips, he hit really bad traffic on the way down and decided if he is going to wait he would rather wait by the beach instead and took a scenic detour

Even though the detour added extra time to the trip, he said he actually felt way better and enjoyed the ride a lot more. That really stuck with me.

When I brought the idea to the team it resonated with all of us because we could all think of similar experiences where the journey ended up being just as important as getting to the destination. We wanted to find a way to give people that same experience intentionally.

What We Learned

We learned a lot throughout the process especially about the Google Maps API and how routing actually works. We had to learn how routing algorithms evaluate different paths and how we could stitch different routes and waypoints together to create a scenic route instead of just finding the fastest one.

We also learned a lot about connecting different parts of an application together from the mobile app and backend to routing, landmarks, narration, and the AI guide.

How We Built It

The main idea behind WayMore was to enhance the passenger experience in autonomous vehicles.

With autonomous driving becoming more common we wanted to look at the experience from the passenger's perspective. If the car is already doing the driving, why should the passenger only care about getting there as fast as possible?

WayMore lets users enter a destination, choose what they want to see, and decide how much extra time they are willing to spend. We then use routing and landmark data to find scenic detours that fit their preferences.

The app also turns the route into more of an experience by showing landmarks along the way, providing narrated information, and allowing the passenger to interact with an AI guide.

Challenges

One of our biggest challenges was that we were mostly all first-time mobile developers. Getting our mobile testing environment set up and finding a reliable way to test changes and troubleshoot problems with live feedback took a lot more work than we expected.

Another challenge was figuring out what we actually needed to prioritize. There were a lot of ideas and features we wanted to add, but we had a limited amount of time. We had to figure out what was important to the core experience and make sure all of our different pieces actually worked together.

In the end, an idea that started from a friend's frustrating drive turned into WayMore a way to make the ride itself more enjoyable, not just the destination.

Built With

Submitted to

ShellHacks 2026

Created by

I engineered and built the entire real-time voice synthesis and AI copilot backend powering WayMore’s autonomous passenger tour experience. To turn scenic driving into an engaging passenger experience, I developed a low-latency proactive narration engine that continuously tracks vehicle kinematics, calculating vector headings, approach speeds, and dynamic ETA windows to cue audio precisely before reaching a landmark. The narration generation leverages an optimized OpenAI pipeline strictly constrained by custom validation guardrails to deliver one to two natural, spoken sentences under twenty-eight words, automatically expanding street abbreviations to ensure clean phonetic pronunciation. I integrated the Speechify Text-to-Speech API with custom SSML prosody rate tuning and expressive emotional styling, backed by a persistent multi-tiered disk-caching layer that enables instant, zero-latency playback and resilience against flaky network conditions.

Alongside automated narration, I developed the live interactive Ask Guide copilot, pairing an OpenAI model with web-grounded retrieval to answer spontaneous passenger questions about sights along the route. I constructed the context injection pipeline that feeds the model real-time telemetry, vehicle direction, recent narration history, and neighboring points of interest, while writing a regex sanitizer that converts raw markdown citations and bracketed annotations into clean, spoken plain text while preserving structured source links for mobile UI cards. To guarantee reliability under API rate limits, I implemented deterministic quota guardrails that gracefully bypass search tools on throttling errors to provide conversational fallbacks without redundant network latency. Finally, I authored an importance-weighted ranking algorithm prioritizing cultural and historic sites over commercial chains, built an internal reverse-proxy architecture to expose unified endpoints for the mobile client, developed an automated drive-replay simulation tool for testing, and implemented a seventy-five-test automated validation suite ensuring robust performance across spatial calculations, audio caching, and conversational grounding.

Alec Borque
Julian Sanchez
Nikolas Enriquez
Isaac Edward