First-person video of people performing real-world tasks across residential, commercial, industrial, retail, and other environments.

Possible configurations:
- Head/chest-mounted or smartphone POV
- Narrated or non-narrated
- Task/action metadata
- Multiple countries and environments
- Raw or annotated
- Custom activities
Real-world product and shelf imagery for product recognition, inventory intelligence, retail automation, and visual search.
Real-world images of:
- Receipts
- Invoices
- Shipping labels
- Product labels
- Packaging
- Filled forms
- Handwriting
Available with transcription and structured metadata.
Video of workers performing manufacturing, maintenance, logistics, assembly, inspection, and equipment-operation tasks.
Driving footage across configurable road, traffic, weather, lighting, geography, and camera conditions.
Conveyors, production lines, warehouses, machinery, logistics operations, etc.

Natural conversations configurable by:
- Language
- Dialect
- Geography
- Demographics
- Environment
- Domain
Available with transcription.
Domain-specific conversational audio for voice agents, ASR, speech analytics, and conversational AI.
Crying, coughing, laughter, breathing, environmental sounds, etc., depending on project requirements and collection feasibility.

Real human interactions with coding agents during:
- Development
- Debugging
- Codebase navigation
- Technical research
- Data analysis
Potential artifacts include conversation history, tool events, terminal activity, Git diffs, screen recording, and metadata.
Existing codebases, repository histories, documentation, and tests for coding-model training and evaluation.
Human evaluation of model responses for:
- Preference
- Correctness
- Grounding
- Instruction adherence
- Hallucination
- Safety/relevance criteria
Multi-step interactions between people and AI agents completing real tasks.

Sequential native CAD revisions showing how engineers move from requirements through design changes, including engineering rationale and revision history.
Real-world architectural, engineering, and construction data capturing how designs evolve from requirements through revisions and final decisions. Datasets can include native BIM/CAD files, sequential design revisions, floor plans, project context, design rationale, change logs, and supporting analyses.
Custom collections can be tailored to workflows in Revit, Rhino, AutoCAD, Tekla, IFC, and other AEC environments for training and evaluating BIM automation, design generation, AI agents, and engineering reasoning systems.

Contact us