Access AI

Inspiration

Trusting that a place is truly accessible shouldn't require deep research, guesswork, or lived experience trial-and-error. Many platforms tag places as "accessible" without context — but for someone using a wheelchair, walker, or with sensory needs, that label can mean wildly different things. Access AI was born to close that trust gap using the power of AI agents and real-world review data.

What it does

Access AI is a conversational travel assistant that helps users discover and navigate cities with their unique accessibility needs in mind.

• It provides a accessibility score of the city as a whole.
• It determines if the specific place has a wheelchair accessibility or not.
• It finds wheelchair accessible place within a city.

How we built it

OpenAI Agent SDK to create a smart, tool-using LLM agent using The Bright Initiative and Nimble datasets.

Challenges we ran into

• Getting the Nimble account set up.
• 30 second timeout in playground.
• Data availability.

Accomplishments that we're proud of

A fully featured POC to enable members of our community with disabilities to find lodging.

What we learned

• When results are shown as a table, it seems to get truncated as default by the AI model.

What's next for Access AI

• Integrate with GPS and local time data to adjust recommendations dynamically (e.g. avoid a crowded museum on free-entry days).
• Cache city guides, accessibility info, and directions for areas with poor connectivity.
• Allow users to rate/review accessibility info, flag incorrect listings, and contribute personal tips.
• Integrate with Alexa, Siri, and Google Assistant for hands-free interaction.
• Co-create and validate features with real communities—e.g. disability rights orgs, eldercare groups, neurodivergent support networks.

Built With

• claude
• databricks
• delta-sharing
• langchain
• mcp
• mlflow
• nimble
• the-bright-initiative
• unitycatalog