The Conja platform is a powerful extended reality workhorse for collecting a broad range of multimodal inputs including accurate spatial measurement, managing generative AI querying and displaying the resultant AI-generated guidance back onto the stage.
All this is achieved using regular mobile devices.
Conja’s novel workflow
To achieve our objectives, Conja has a formidable tech stack:
Computer Vision AI integrations and capabilities
Machine Learning, Models, AI On Device Integrations, Google ML, Apple CoreML, Open Source Models, UltraLytics, Meta Vision Models.
Capabilities: Object Detection, Body Tracking, Body and Pose Tracking, NSFW Visual Safety, Segmentation, Background Removal, Zero Shot Classification, YOLO, Text and Barcode recognition.
Roadmap: Partner Vision Models for Environment Feature Detection, Enterprise Equipment Inventory detection.
Spatial Computing Technologies
iOS, iPhone, iPad
ARKit, RealityKit
OpenUSD, Reality Files
Multi Model Support
Authoring Enabled In Spatial Computing Stage (No Unity Dependency)
Augmented Reality support for Images, Video, Text, 360 Video, Holograms, Spatial Audio, Digital Twins and Interactive Models
Lidar Measurement Tools
NeRF mapping and rendering
Local, Remote Cloud Galleries
iCloud, Google Drive, Photo Gallery, Google Cloud, Firebase
Generative AI Assets
Roadmap: NVidia Omniverse Drive Integration, NVidia Deep Search, NVidia Omniverse APIs, Digital Twins with AI integrations, Environmental models using Neural Radiance Fields, Vision based surface detection, Advanced Anchoring
Computer Vision Cloud AI Integrations
Google Gemini, Google Vision, Landing Lens
Roadmap pipeline: Google Vertex Pipelines, NVidia Vision Pipelines, Roboflow, Enterprise Computer Vision Models
Generative AI XR Integrations
GenAI: Google Gemini, Google ImageGen
Roadmap: Google Veo, Stability AI (3D Generation), Luma (Nerf GenAI), Luma Genie (3d), OpenAI GPT Vision (Multi-modal XRxAI)
NVidia Nemo, Picasso, Omniverse NIM Pipelines, Multi-modal chaining pipelines
Key considerations in our development strategy:
We are trainers first, so what we build has to be guaranteed accurate, compliant, valuable and measurable against benchmarks.
We are not looking to replace the skilled Ot. We are expanding their reach, co-piloting to improve productiveness and incorporating documentation to relieve unnecessary stress points at work.
We are mapping processes based on a combination of professional lived experience, industry best practice and legislative demands, so our demands on AI are task specific.
To be effective, Conja must manage diverse input, it must be able guide the user through the process. It must also be able to output AI driven guidance as if demonstrated by an instructor, therefore rich 3D multimedia output is the preferred method.
Further considerations
Conja must be agile and available on demand
Preference of Mobile Hardware Not Goggles • Mixed Reality Not Virtual Reality • Mobile Spatial Computing Interfaces Not Cloud • Look through the App Not at the App • Workers in the home Not in the Office • Train On the Field Not in Seminars • Workers on the move Not tethered to the Desktop