-
-
Dashboard
-
Inspiration: Industrial manufacturers lose approximately $50 billion every year due to unplanned downtime. We realized that traditional maintenance is either Reactive—fixing things only after they break—or Preventative, which often replaces perfectly healthy parts based on a rigid schedule. We were inspired to build FactoryAI because operators currently lack a unified system that can correlate sensor telemetry, visual data, and supply chain logistics in real-time.
What it does: FactoryAI is an autonomous orchestration engine that moves the factory floor from simply "Alerting" to "Acting". Predictive Analysis: It monitors 10Hz telemetry to predict potential equipment failures up to 48 hours in advance. Autonomous Logistics: The system automatically checks inventory, orders parts from vendors, and schedules the necessary technicians. Visual Verification: It utilizes computer vision to confirm repairs and ensure safety compliance.
How we built it: We built a Multi-Agent architecture powered by Gemini 3.0 Flash. The Sentinel: This agent handles anomaly detection by monitoring edge sensors for vibration, temperature, and pressure. The Strategist: This agent manages procurement by checking stock and searching for the best vendor based on lead times and cost. Reasoning Engine: We used JSON Mode with gemini-3-flash-preview to force strict, deterministic outputs that control the application state. Vision Nexus: We integrated multimodal analysis where video frames are sent to Gemini to act as a "second pair of eyes" for validating repairs.
Challenges we ran into: The biggest challenge was ensuring Deterministic Reliability. In an industrial environment, the AI cannot be "creative" with machine status. We solved this by using structured prompts to force the model to follow a strict Chain of Thought process. For example, when Machine M-142 spiked to 92°C, the AI had to logically conclude that a "Cheap" vendor with a 72-hour lead time was unacceptable because the estimated failure time was only 36 hours.
Accomplishments that we're proud of: We are proud of creating a system of Checks and Balances through Agent Personas. By instantiating roles like the "Risk-Weighted Economist" and the "Skeptical Inspector," we ensured that every autonomous decision is verified by distinct specialized logic before execution.
What we learned: We learned that Gemini 3.0 is more than just a chatbot; it is a sophisticated Reasoning Engine. It can handle the math required to save a factory's bottom line. By calculating that a $150 "Fast" vendor was a better choice than an $80 "Cheap" vendor to avoid a total shutdown, we proved the ROI of autonomous AI.
What's next for FactoryAI: The next step is scaling our data ingestion from 10Hz to 100Hz to capture even more granular edge sensor data. We also plan to expand the Human-in-the-Loop verification, allowing for more complex repairs to be guided and verified by Gemini in real-time.
Log in or sign up for Devpost to join the conversation.