DAISAP

• 

  Giving final output after running through a day's worth of planning.

Inspiration

DAISAP was inspired by the challenge of navigating the vast session landscape at the Databricks Data + AI Summit. With hundreds of sessions spanning AI, ML, engineering, governance, and industry use cases, we wanted to create a personalized, intelligent scheduling assistant that helps attendees cut through the noise and walk away with a tailored agenda aligned to their professional goals. Our vision was to go beyond static filters and use AI to match users with high-value content — and to make it memorable and easy.

What it does

DAISAP is a personalized AI scheduling agent that helps Databricks Summit attendees build an optimal session agenda. It guides users through an interactive intake flow, learns their preferences (e.g., technologies, tracks, experience level), and then recommends the best sessions per hour, per day — avoiding overlaps, preserving breaks, and offering backup picks.

How we built it

We started by uploading the official Summit session data into the Databricks Unity Catalog as a managed table. From there, we enhanced it with intelligent tooling — including a custom function tool for session filtering and a vector search index tool to semantically match user preferences with session content. The core intelligence of DAISAP was built through iterative prompt engineering.

Challenges we ran into

We had trouble creating functions to insert data, so we just save it in session memory. Had trouble showing function calls, but not showing the results. example: {"_tool_reasoning":"To provide session options for the next time block, we need to retrieve sessions for the 10:00 AM time block on Tuesday.","p_day":"Tuesday","p_hour":"10:00 AM"}

Accomplishments that we're proud of

Persevered through the challenges we faced. Collaboration and putting ourselves out there for the hackathon to meet new people and work together.

What we learned

We learned that functions are hard to use for merges. Crucial prompt engineering and training of the agent was an iterative process.

What's next for DAISAP

Future enhancements could include persisting the data to allow users to resume any time, and adding a leave feedback dialog. Converting it into a chatbot to register your session picks as well.

Built With

• databricks