Knolia.ai: From Passive Video to Interactive Knowledge

Conference knowledge often disappears after an event ends—or sits unwatched on YouTube. Anurag Duddu (MDes 2025) saw an opportunity: what if conferences could stay interactive months or even years later?

Built for the 2025 Shapeshift Summit at the Institute of Design at Illinois Tech, Knolia.ai serves as a knowledge liaison between Summit attendees and the insights shared across three days exploring responsible AI. Using speech-to-text processing, speaker diarization, and retrieval-augmented generation (RAG), the platform transforms passive recordings into an interactive knowledge layer—allowing participants to query summit content semantically, receive grounded responses with exact citations, and experience talks in their own language through live dubbing in speakers’ original voices.

While created specifically for Shapeshift, this approach could be applied to other summits, conferences, and convenings—building community and knowledge long after in-person events end.

Beyond Passive Playback

Traditional conference recordings offer flat experiences. YouTube playback with keyword search requires exact matches and forces manual scrubbing through hours of video. Even when transcripts are available, they’re difficult to query effectively.

How can we go beyond passive playback?

Generic large language models make everything available—but risk pulling from unreliable sources with unknown authority. Participants can’t be certain whether responses reflect what was actually said at the Summit.

Knolia takes a different approach: a closed-source system that allows participants to engage with and query only trusted Summit content. Events stay interactive long after they end.

Knolia emerged directly from Shapeshift’s focus: Responsible AI. The aim of the Summit was to move AI conversations from Silicon Valley to the rest of the world, and from technology to human impact.

Shapeshift spotlighted responsible approaches to AI—technology designed to enhance human capability rather than replace it, and systems built to serve people rather than exploit them. Anurag’s work embodies these principles. Rather than building another general-purpose AI tool, he created a specialized system grounded entirely in verified Summit content—demonstrating how emerging technologies can be applied responsibly to build community and preserve knowledge.

The fundamental shift Knolia enables becomes clear when comparing traditional approaches:

A participant might ask: “What did Ryan Powell from Waymo say about designing for trust in urban systems?” Knolia returns grounded responses citing specific sessions with direct links to those moments—transforming how people extract value from conference content on their own terms, months or years later.

For Shapeshift Summit registrants with platform access:

1. Visit knolia.ai and log in via Google Workspace with your Summit credentials
2. Ask natural language questions about any session, speaker, or topic discussed across the three-day Summit
3. Receive grounded responses with exact citations pointing to specific moments in specific sessions
4. Navigate directly to moments via word-level timestamps that take you to relevant content
5. Experience content in your language through live dubbing that retains speakers’ original voices
6. Explore auto-generated highlights and session summaries for quick orientation

The platform processes summit content through integrated layers: speech-to-text with speaker diarization identifying who said what, embedding with word-level timestamps, semantic search using RAG technology to ensure responses stay grounded in actual content, and multi-language support rendering talks in users’ preferred languages.

More than 17 percent of Shapeshift registrants have used the prototype platform, and many of them keep coming back. Nearly all returned multiple times over the following months to interact with Summit content—demonstrating sustained engagement with conference knowledge long after the May 2025 event ended.

This engagement pattern reveals an unmet need. Conference recordings typically see a brief spike in views during the first week, then virtually no activity. Knolia demonstrates what becomes possible when you transform passive recordings into interactive knowledge layers: people find ongoing value in event content because the format finally supports how they actually want to access knowledge.

While built as a prototype exploring what AI can do for event knowledge, Knolia demonstrates how responsible AI can keep communities connected to the ideas that matter. The platform remains live at Knolia.ai for Shapeshift registrants.

The approach is replicable: other summits, conferences, and convenings could apply these same methods to build community and preserve knowledge long after in-person events end—creating interactive knowledge layers that serve participants for months and years.

Anurag Duddu (MDes 2025)

Anijo Mathew (Shapeshift Co-Host)

Albert Shum (Advisor to the Dean; Shapeshift Co-Host)