Gemba Robotics provides a factory intelligence platform that creates a digital twin of a manufacturing plant from a single drone flight, using on‑device computer vision to generate geometry‑accurate 3D models without CAD data. The twin feeds a high‑performance discrete‑event simulator, enabling plant managers to run thousands of what‑if scenarios, identify bottlenecks, and perform sensitivity analysis via plain‑language queries. Data are encrypted and access‑controlled for privacy.
Funding
Funding not disclosed
Founders
Product
Problem
Manufacturing plants often rely on tacit knowledge and manual observations to assess workflow, leading to invisible congestion, inaccurate capacity planning, and outdated models that fail as layouts or volumes change. Without precise, up‑to‑date spatial data, operators cannot reliably identify bottlenecks or predict the impact of operational changes.
Solution
Gemba Robotics delivers an end‑to‑end factory intelligence platform that transforms a single drone flight into a living digital twin of the facility. On‑device computer‑vision pipelines anonymize video streams in real time, then a zero‑shot reconstruction engine generates a geometry‑accurate 3D model without any CAD input. The model feeds directly into a discrete‑event simulation engine that can run tens of thousands of scenarios, providing statistically grounded insights into throughput, congestion, and capacity limits. Users interact with the simulation through a plain‑language query interface that supports sensitivity analysis and “what‑if” planning, enabling continuous model updates as the plant evolves. All data are encrypted in transit and stored with role‑based access controls, ensuring privacy and compliance.
Target Audience
Primary customers are plant managers, operations engineers, and industrial analytics teams at mid‑size to large manufacturing facilities seeking data‑driven workflow optimization and risk‑aware layout planning.
Features
- Drone‑based capture covering an entire facility in a single, non‑disruptive flight
- On‑device AI segmentation that classifies and pixelates people, forklifts, and equipment in real time for privacy‑first data handling
- Zero‑config 3D reconstruction pipeline that derives geometry, aisle widths, and spatial relationships directly from video footage
- Traffic‑lane overlay that projects observed vehicle paths onto the digital twin for accurate flow mapping
- Simulation‑ready output that integrates seamlessly with a high‑performance discrete‑event simulator (≥50,000 events per scenario)
- Plain‑language interrogation engine for bottleneck identification, assumption testing, and optimization recommendations
- Sensitivity analysis tools that rank model parameters by impact, guiding data‑collection priorities
- Optional fixed‑camera augmentation to enrich multi‑shift traffic data and improve model fidelity