Experiments

AI Lab

Exploring how AI fits into creative production — not as a replacement for people, but as a multiplier for teams. These are experiments, prototypes, and explorations at the intersection of AI and the tools creative teams actually use.

In Progress

LLM-Assisted Pipeline Debugging

Exploring how large language models can parse production logs, identify failure patterns, and suggest fixes for pipeline errors — reducing the feedback loop from hours to seconds.

LLMPipelineAutomation
Prototype

Generative Asset Descriptions

Using AI to auto-generate metadata and descriptions for production assets, making search and discovery across large asset libraries practical for the first time.

NLPMetadataProduction
Concept

Conversational Project Standup

A natural language interface for daily standups — team members talk to it, it structures updates, flags blockers, and routes information to the right people.

AgentsTeam OpsLLM
Exploration

AI-Driven Shot Review

Automating first-pass QC on renders — catching technical errors like missing AOVs, color space issues, and naming violations before a human ever looks at it.

Computer VisionQCVFX
Prototype

Code Generation for DCC Scripts

Fine-tuning and prompt engineering for generating Maya, Houdini, and Nuke scripts from natural language descriptions of artist workflows.

Code GenDCCPython