All In Vegas

Inspiration

All In Vegas was born from the chaos of navigating conventions in a city that never stops moving. Las Vegas hosts some of the most intense, high-stakes conventions on the planet with sprawling venues, packed schedules, and dozens of sessions happening at once, and yet attendees were still left juggling PDFs, scattered venue maps, and word-of-mouth to figure out where to be and when. We wanted to build the platform we wished existed: one that puts discovery, navigation, and scheduling in the same place, and actually understands what you're looking for. For operators, the frustration was equally real. There was no good way to create a presence, place their venues on a map, and reach the right audience in one move. Bringing generative AI into the loop let us go further than a glorified event calendar. It meant the platform could match what's happening across the floor to what a specific attendee actually cares about and help them build a schedule that makes sense, in minutes, not hours. Vegas moves fast. We built something that can keep up.

What it does

All In Vegas serves two kinds of users: attendees who want to discover and plan, and operators who want to create and manage. For attendees, the platform lets you browse a catalog of conventions and venues, filter by category, inspect locations on a live map, and build a personal schedule through the calendar view. An AI chat assistant sits alongside the experience where you can ask what's worth attending based on your interests, and it surfaces curated recommendations with context on why each one fits. For operators, a multi-step Event Builder lets them create conventions and subevents, describe their venues, and pin them directly on a map, giving attendees accurate, visual information from day one.

How we built it

We focused on modular, testable components and rapid iteration: Frontend: React + TypeScript with Vite for fast dev feedback; react-hook-form and Zod for robust forms; Tailwind + shadcn for consistent UI. Maps & UX: @react-google-maps/api to visualize venues and let users pick or inspect locations. AI: An RTK Query-backed AI service that proxies prompts to a model (server-side) and returns curated text responses or structured suggestions. Backend: Four Java 21 AWS Lambda microservices handle events, users, venues, and AI separately — each with its own responsibility and independently deployable. DynamoDB backs persistence, Cognito handles authentication, and S3 handles file storage. Architecture: Feature-based structure so event discovery, schedule building, and venue management evolve independently and can be tested in isolation.

Challenges we ran into

Dual-role UX: Designing a single app that serves both attendees browsing events and operators building them required careful route guarding, conditional UI, and a shared component library that worked cleanly across both contexts without bloating either flow. Output consistency: Early model outputs varied in structure (lists vs. narrative). We iterated on prompt templates and post-processing to normalize responses for the UI.

Accomplishments that we're proud of

An Event Builder that lets operators create a convention, add venues, and place each one on a map through a guided multi-step flow which helps turn a complex setup into something approachable. An AI chat experience that goes beyond keyword search: it understands free-text questions about the event ecosystem and returns recommendations grounded in real platform data. Modular APIs and frontend features that let us toggle between narrative recommendations and structured event lists without breaking the UX.

What we learned

Two-sided products need discipline early: The attendee and operator experiences share UI infrastructure but diverge fast. Establishing clear role-based routing and component boundaries from the start saved us from painful refactors later. Combine strengths: AI is great at high-level curation, but pairing it with deterministic data (venue records, maps, etc.) makes outputs actionable and trustworthy. Confidence-first UX: Showing confidence levels and provenance (where a recommendation came from) increases user trust and makes edge cases easier to handle. Build for iteration: Prompt engineering and post-processing rules are living code keep them in version control and test suites.

What's next for All In Vegas

Operator analytics: Operators can create events and place venues, but they have no visibility into how attendees are engaging with their listings (like through views, saves, AI mentions). A lightweight dashboard would give organizers signal they can actually use. Smarter AI with session context: The current chat assistant answers individual prompts well. The next version should remember what you've already saved, what you've dismissed, and what you've asked before so recommendations get sharper the more you use it, not just the more you prompt it. Real-time event data: Vegas doesn't stop changing. The enrichment layer needs live feeds, whether through operator updates, webhooks, or integrations with platforms like Eventbrite, so the AI is always recommending against current listings, not stale ones. Mobile-first experience: Conventions are a phone-in-hand environment. A Progressive Web App or native mobile wrapper would put the map, schedule, and AI assistant exactly where attendees actually need them which is on the floor, in real time.