The Inspiration: Bridging the Gap Between Tech and Cinema My lifelong love for film isn't just about winning movie trivia; it’s rooted in a deep fascination with the mechanics of the creative process and how the magic is actually made behind the scenes. Bridging my professional background in tech with the realities of the film industry—sparked by candid conversations with my family members who work as producers—uncovered a massive operational bottleneck. The industry is plagued by non-standard tools. Freelancers and artists all have their own unique workflows, which becomes a logistical nightmare when dealing with storyboards. Translating hand-drawn sketches, custom PDFs, and iPad doodles into a cohesive, structured format requires countless hours of grueling manual labor from production crews. I saw an opportunity to let technology adapt to the artists, rather than forcing the artists to adapt to the technology.

The Solution: Project Rosetta Board To solve this, I built Project Rosetta Board (The Universal Storyboard Ingestor). It is an autonomous, multi-agent pipeline built natively on the Gemini Enterprise Agent Platform designed to structure the chaos of pre-production. Here is exactly what the system does:Multimodal Ingestion: The system accepts any unstructured storyboard format and uses Gemini 1.5 Pro to "look" at the raw sketches, extracting both explicit text and visually implied metadata, such as unwritten camera angles or props. Nomenclature

Normalization: An ADK-orchestrated Schema Agent cross-references the extracted data against a master studio taxonomy, instantly mapping messy freelancer jargon into standardized production terms. Real-

Time Data Storage: The cleaned, structured data is sent through the official ClickHouse MCP server and committed to a ClickHouse Cloud database.

Dynamic Output: What starts as a messy pile of disparate sketches becomes a centralized, instantly queryable database that Assistant Directors and Prop Masters can use to seamlessly generate daily schedules and budget estimates. Ultimately, you built a universal translator that preserves the director's visual vision while delivering the strict, structured data the production crew needs to actually shoot the movie.

Built With

  • clickhouse-cloud
  • clickhouse-connect
  • clickhouse-mcp-server
  • fastapi
  • gemini-2.5-flash
  • gemini-2.5-pro
  • google-agents-cli
  • google-cloud
  • google-cloud-adk
  • google-cloud-run
  • google-cloud-secret-manager
  • google-cloud-vertex-ai
  • hatchling
  • model-context-protocol-(mcp)
  • pandas
  • parallel-api
  • pydantic
  • pytest
  • python
  • reportlab
  • sqlite
  • streamlit
  • uv
  • uvicorn
  • vertex-ai-agent-engine
Share this project:

Updates

Submission history