Summary, problem, and solution

Dark factories are fully automated factories that operate without on-site human labor, relying on automation, AI, and robotics. They already exist in countries such as Japan, China, and the Netherlands.

As advances in AI and robotics make dark factories more common, humans are still needed in the loop as observers and maintainers of these complex systems. Even with extensive sensors and intelligence, automated systems can miss issues—such as unexpected process failures or gradual declines in output quality.

For this reason, human oversight remains essential. Imagine a small, warehouse-sized factory where a single person walks through once a day to verify that everything is functioning as expected. We have built a platform that allows this person to log issues they observe and ask deep, non-obvious questions about the factory—questions that cannot be answered by telemetry data alone.

Technical

We use a voice agent built on ElevenLabs and OpenAI APIs to hold natural conversations with the human operator. In addition to standard factory telemetry logs, we maintain a separate database of issues reported by the operator.

By combining these data sources with dynamically generated code executed in Daytona, the system can answer non-obvious, complex questions. As shown in our demo, Daytona generates and runs new code to build the required graphs on demand.

Sponsor Tools

• Daytona (run multiple python jobs at once)
• ElevenLabs (text-to-speech, and speech-to-text for natural conversation)

Built With

• daytona
• openai
• python