AI Archive Enrichment for MediaStorm

The Archive

Brian Storm founded MediaStorm in 2005. Over two decades, the studio has produced 350+ documentary stories with collaborators including CNN, National Geographic, the Wall Street Journal, and the UN Foundation. Character-driven documentaries about displacement, conservation, conflict, human rights.

I've followed Brian's work since I started in media. This was a project I wanted to build.

The Opportunity

MediaStorm's archive is well-catalogued — titles, descriptions, full transcripts with speaker attribution. But 350 stories spanning two decades hold more than keyword search can reach. The goal was to give the archive a deeper layer of access — searchable by theme, tone, narrative structure, not just words that happen to appear in a title.

What We Built

A pipeline that ingests the entire published archive, enriches it with AI, and makes every story findable by meaning.

Each story runs through Gemini to add new layers of structure — topics, themes, named entities, emotional tone, narrative structure. Multimodal embeddings index both the text and the visual content — poster frames, thumbnails — so the search understands what a story looks like, not just what it says.

A story about a Kenyan wildlife caretaker gets tagged with conservation, Kenya, Ol Pejeta Conservancy, and a "hopeful, bittersweet" tone. That structure didn't exist before.

Everything gets embedded and indexed for hybrid retrieval — keyword and semantic search fused together, then reranked. A query router detects whether you're searching for a person, a theme, a time period, or an award, and adjusts the retrieval strategy accordingly.

"Show me stories about families affected by war" returns relevant films with themes, contributors, and awards — not keyword matches. The system understands displacement, not just the word "war."

What This Opens Up

Every media organization sits on an archive. Decades of footage, transcripts, reporting — catalogued well enough to store, but not well enough to search by meaning. AI changes that equation. It turns archives into knowledge bases — searchable, structured, and accessible in ways that weren't possible before. The librarian that never forgets.