StudioPilot

• 

Inspiration

Film and television productions are highly interconnected. A single unavailable actor, location, or piece of equipment can create a chain of schedule conflicts that is difficult to resolve manually.

We built StudioPilot to help production teams recover from these disruptions faster. Instead of only identifying what went wrong, StudioPilot analyzes the production state, evaluates conflicts, proposes schedule changes, and lets the producer approve the recovery plan before anything is applied.

What it does

StudioPilot is an AI-powered production recovery assistant.

A production schedule is loaded into the system with scenes, cast, locations, equipment, and availability constraints. StudioPilot can:

• Detect conflicts across the production schedule
• Identify affected scenes and shooting days
• Classify conflicts by severity
• Generate an AI-assisted recovery plan using Gemini
• Explain proposed schedule changes and trade-offs
• Keep a human producer in control through an approval step
• Apply approved changes to the production schedule
• Show the resulting schedule and confirm when conflicts have been resolved

The core workflow is:

Production → Detect Conflicts → Generate Recovery Plan → Human Approval → Apply Changes

How we built it

StudioPilot is a web application with a Python/FastAPI backend and a browser-based frontend.

The backend manages the production state, conflict detection, recovery planning, approval workflow, and schedule mutations. Gemini provides the AI reasoning used to generate recovery recommendations.

The application is deployed on Google Cloud Run, with the Gemini API credential managed securely through Google Cloud infrastructure.

IBM Bob was used throughout the development process to build, inspect, debug, test, and prepare the project for deployment.

The system uses a structured production model so that AI recommendations are constrained by actual cast, location, equipment, scene, and scheduling data rather than being treated as free-form suggestions.

Challenges we ran into

One of the biggest challenges was making AI recommendations useful while keeping the production rules deterministic.

A recovery plan cannot simply sound reasonable; proposed changes need to respect the constraints already present in the production.

We therefore separated constraint validation from AI reasoning. The application verifies production constraints while Gemini helps determine practical recovery options.

Deployment was another challenge. We needed the same workflow that worked locally to operate reliably on Google Cloud while keeping the Gemini API credential secure.

Accomplishments that we're proud of

We are proud of building an end-to-end production recovery workflow rather than just an AI chat interface.

StudioPilot can take a production with real scheduling conflicts, identify the problems, generate a proposed recovery plan, place the producer in the approval loop, and apply the approved changes to the schedule.

The final workflow demonstrates how AI can support production decision-making while keeping humans responsible for important operational decisions.

What we learned

We learned that effective agentic systems need more than a capable AI model. The surrounding application needs clear state, constraints, tools, validation, and human control.

We also learned that AI is most useful in this workflow when it helps a producer make a difficult decision rather than simply generating text.

Building and deploying StudioPilot also gave us practical experience with Gemini, Google Cloud deployment, secure configuration, production workflows, and using IBM Bob throughout development.

What's next for StudioPilot

The current version focuses on production schedule recovery.

Future versions could connect to live production-management systems, incorporate real-time availability changes, provide deeper cost and resource optimization, and support larger multi-production workflows.

The long-term goal is to make StudioPilot an intelligent production operations layer that helps teams respond to disruption before it becomes a production crisis.

Built With

• css
• fastapi
• gemini
• google-cloud
• googlecloudrun
• html
• ibmbob
• python