An AI-powered LED wall for Archive, built in three weeks
Archive needed a showstopper for its booth at The Lead Summit: a wall that proves their platform sees what is happening on social media, live. We built it in three weeks. AI watched fresh TikTok and Reels about each brand and wrote the brand's story on the wall as the conference went on.
All case studies- Client
- Archive, an AI platform that tracks creator content and brand mentions across TikTok, Instagram and YouTube
- Event
- The Lead Summit, May 2025
- Service
- AI integration, real-time content pipeline, event software
- Built on
- Django, Celery, WebSockets, Gemini via Braintrust, Archive's own data webhooks
- Timeline
- about three weeks, from first ticket to the show floor
The situation
Archive sells one promise: it captures every mention of a brand on social media, even the ones without a tag.
A conference booth is the hardest place to prove that. Visitors give you a few seconds. Slides do not convince anyone.
So Archive wanted the platform itself on a giant LED wall, working live, in front of the people walking by. With a fixed date and no room to slip.
What we built
A tablet controller and an LED wall.
On the tablet, the team picks a brand. The wall shows that brand's top creator videos, its numbers and a story about what is happening with the brand on social right now.
The story is not written by a person. AI writes it from real, fresh videos.
How the AI works
Archive's platform sends us webhooks about new posts, reach updates, transcripts and sentiment. We keep only what is safe to put on a wall in public: Instagram Reels and TikTok, brand-safe, positive sentiment, single-video posts, sorted by views. Then two AI steps run for each brand.
- 01
The model watches each video and writes a short summary from the video itself, its transcript and its metadata.
- 02
It writes the next chapter of the brand's story from those summaries, the brand's details, the earlier chapters and the post statistics.
Each brand gets a playlist of its top 20 stories. The numbers on screen (views, posts, creators, likes, comments, earned media value) feed the story too. Regeneration runs in the background, so the wall never waits for the model.
What broke, and how we fixed it
- Video would not go through. The AI proxy Archive uses did not accept video. Archive chose Gemini 2.0 Flash Lite, and we switched the pipeline to it.
- Some videos arrived with the wrong file type from the CDN, and the model rejected them. We now flag and skip them instead of stalling.
- A newer model version started returning empty stories. We rolled back to the stable one.
- Some calls hung for up to two minutes. We added monitoring, and any video without a summary is skipped, so one bad file never blocks a brand.
- Hundreds of posts from different brands shared the same shop ID. We identify the brand by its mentions instead.
Results
The wall ran live at The Lead Summit.
From scratch, a brand's full story was generated in about five minutes. With summaries already in place, a refresh took about 40 seconds.
In Paul's words, recorded at the booth: "They helped us build this in three weeks. It works perfectly."
“We worked with the team over at Owlab who helped us build this incredible technology. They helped us build this in three weeks. It works perfectly. Couldn't have done it without you guys.”
What happened next
Archive kept working with us: an AI LinkedIn outreach flow on n8n, a Slack to Linear support bot and design work.
Have an event, a launch or a demo that has to work live?
We start with a fixed-price pilot from $1,000.
Book a call about your processHave an event, a launch or a demo that has to work live?
Tell us in two lines. We'll reply with how an agent would take it over and what the first step would cost.
- A fixed-price first step, from $1,000
- Half at the start, half after we check it on your data
- NDA on request
- [email protected]Vitaliy, CEO
- Vesivärava tn 50-201, 10152 Tallinn, EstoniaHead and development office