Factory Shift is an interactive scenario-based simulation for advanced manufacturing and industrial tech programs. Instead of coding models, students step into real-world production environments to diagnose model drift, false passes, sensor blindness, and operator workarounds when confident AI systems fail.
Learners take on the role of line technicians or quality specialists facing realistic production anomalies (e.g., machine vision defect inspection, vibration alerts).
Instead of trusting raw confidence scores, learners inspect the data pipeline: camera lighting, corrupted training sets, edge sensors, and physical bottlenecks.
Learners cross-reference algorithmic alerts with operator feedback to discover why workers bypass or ignore AI flags.
Students propose operational solutions—from re-labeling edge cases to establishing human-in-the-loop escalation protocols.
Data Integrity & Sensor Drift
When a machine-vision inspection model approves out-of-spec welds or parts, learners diagnose why the confidence score was high. They practice isolating lighting shifts, detecting tainted test splits, and auditing camera focal drift before blaming the underlying model.
Human-in-the-Loop Workflow
An AI predictive maintenance model issues 50 daily warnings that operators routinely silence. Learners interview virtual operators to identify why the alert was ignored, discover unwritten shop-floor workarounds, and redesign the handoff protocol so every alert has a clear human owner.
Root-Cause Triage & Governance
Learners build disciplined post-incident reflexes: isolating what the model actually ingested, documenting system blind spots, and determining whether to retrain, recalibrate physical hardware, or change operational tolerances.
Drop scenario labs into existing mechatronics, automation, and industrial tech courses. Students practice diagnosing AI failures beside PLCs, sensors, and quality systems—without needing a data-science track.
Give apprentices and incumbent workers a safe place to rehearse smart-factory judgment: when to trust a model, when to escalate, and how to document what the line actually saw.
Upskill operators, quality leads, and supervisors on human-in-the-loop habits before a new vision or predictive-maintenance rollout hits the floor.
A short explainer on teaching investigation habits for shop-floor AI — data, process, and people.