I am Sara
A real-time AI avatar on the web and four centrally controlled interactive exhibition displays.
For I am SARA, we built iamsara.io and four interactive 9:16 displays for Data Expo 2026. SARA holds live conversations, automatically follows the visitor's language, searches a controlled knowledge base with 117 exhibitors and 171 programme sessions, draws routes on the floor plan and remains manageable from one privacy-safe console.
What it does
Face-to-face, without touch
Visitors talk hands-free with SARA on the portrait display. Local person detection starts an anonymous conversation when someone steps onto the floor position; camera frames are never uploaded or stored.
Multilingual from the first sentence
Dutch, English, German and French have reviewed starting flows. In automatic mode, the conversation can also follow another clearly spoken language when supported by the speech providers.
The entire event as a knowledge base
SARA searches controlled local copies of 117 official exhibitor profiles and 171 programme sessions, plus opening hours, halls, stands, travel and practical information.
From answer to route
After explaining a session or exhibitor, SARA draws the route from the visitor's current station across the official event floor plan.
Plan videos from operations
Uploading, previewing, ordering and enabling or disabling each clip happens centrally. Kiosks cache approved media locally, pause for a visitor and resume afterwards.
Four displays, one live console
Online status, response time, actions, badges, privacy-filtered conversation snippets and current discussion topics come together in one operational view.
ReSpeaker only listens to the spot
The XVF3800 sends only the sound direction through a local USB bridge. An 85–95 degree listening zone attenuates aisle noise; profiles and volume are remotely adjustable without storing audio.
Badge, pitch and follow-up with consent
A badge is only scanned for a selected action, details are confirmed by voice and a pitch lasts no more than one minute. Marketing consent remains separate and off by default.
Interactive exhibition floor
Four SARAs. One live exhibition floor.
For Data Expo 2026, we translated the real-time AI avatar from iamsara.io into four autonomous exhibition displays. Visitors talk hands-free, immediately find an exhibitor or session, and see the route on screen. The I am SARA team controls the entire experience from one live operations console.
Test view · sample data
- 4SARA displays
- one central platform
- 117Exhibitors
- controlled profiles
- 171Sessions
- programme knowledge base
- 85–95°Listening zone
- directional ReSpeaker audio
The visitor journey
Not a website on a large screen, but a real exhibition guide.
Every step is designed for a busy hall: touch-free, instantly understandable, and ready for the next visitor after every conversation.
- 01
Step closer
Local person detection recognises when someone stands on the floor position and starts an anonymous conversation. Camera frames and tracking data never leave the display.
- 02
Just talk
Dutch, English, German and French are fully reviewed starting flows. In automatic mode, SARA can also follow another clearly spoken language when supported by the speech providers.
- 03
Find what matters
SARA searches controlled local copies of the official programme and exhibitor directory, including times, halls, stands and practical event information.
- 04
Walk the right way
After answering, SARA draws the route from the visitor’s current display. A badge scan, pitch or follow-up only begins after an explicit choice.
Knowledge base & route
From a loose question to a place on the floor plan.
Event knowledge is not loose text inside a prompt. Exhibitors, programme sessions, opening hours, halls and routes are structured sources. SARA can explain what is relevant first, then immediately show where the visitor needs to go.
Interface test
Interface test
Language without friction
Four fixed entry points. Then the conversation follows the visitor.
The opening choice makes the first seconds predictable in Dutch, English, German or French. Visitors who switch languages do not need to return to a menu: the real-time conversation partner moves with them.
4 reviewed starting flows · automatic language recognition during conversation
Remote video
The playlist is planned remotely and played locally.
The team can upload, preview, order and enable or disable films for every kiosk. Each display downloads approved media once into its local cache. The playlist runs smoothly when the station is idle, pauses immediately for a visitor, then resumes afterwards.
Live visibility
Four displays controlled from one console.
The console shows whether each station is online, which action is running and how quickly it responds. Conversation snippets pass through a privacy filter first; full transcripts, camera images and audio do not belong in this view. Film sound, volume, listening profile, audio filtering and optional gesture reactions can all be adjusted per display from the same console.
Test view · sample dataPurpose-built audio
SARA listens to the spot, not the entire hall.
A ReSpeaker Flex with XVF3800 aims the listening zone at the floor position in front of the display. A local, read-only USB bridge supplies only the sound direction. Voices from the surrounding aisle are attenuated; neither spoken words nor raw audio are stored by the bridge or sent to the console.
If the hardware bridge is unavailable, the standard microphone path keeps working. The team can adjust the filter profile, low-cut, attenuation and volume remotely without walking to each display.
Direction only · no audio storagePrivacy as system behaviour
Operational insight without a surveillance console.
Online status, response time, selected actions, anonymised conversation summaries and current topics help the event crew. Names, contact details and badge data stay out of the topic cloud; marketing consent remains separate from functional follow-up and is off by default.
How we built it
Alongside the public website, we built a dedicated exhibition application for four portrait displays. The avatar, knowledge base, video and central controls form one recognisable experience.
Information about exhibitors, the programme, opening hours and routes is checked in advance and made available locally. This keeps SARA fast and reliable, even on a busy exhibition floor.
From one central control point, the team can prepare, order and manage videos for each display. Approved media is stored locally and the display switches immediately when a visitor starts a conversation.
The console shows only what is needed to operate the four displays. Person detection and directional audio remain local; camera images, conversations and raw audio are never sent to the dashboard.
Why we built it
An exhibition app is not a website on a larger screen. In a noisy, brightly lit hall, the experience must start without touch, follow one voice from the crowd, provide immediately useful answers and then reset cleanly for the next visitor.
Four personal conversations at once only work when content, audio and status remain centrally manageable. That makes the operations console the working heart behind the visible SARA displays, not an afterthought.
Need something like this in your company?