Artificial Intelligence

AI that solves real tasks. We apply voice, vision and agentic workflows to deliver measurable outcomes. From pilot to production, dilium builds systems your team can operate and scale.

What we deliver

• Neural voice for dynamic storytelling and content localization using cloned voiceprints of professional voice actors - NDA-protected campaigns in retail.
• Document AI to classify, extract and route documents into custom CMS workflows - deployed with Gabetti.
• Agent AI for tailored pipelines and automations orchestrated with tools like n8n and Microsoft Copilot Studio.
• Vision QA and inspection to detect scratches and geometry issues - Deeteck project.
• Realtime tracking with YOLO in camera streams - Spatial Plan project with Apple enterprise APIs.
• Photogrammetry to generate high-fidelity 3D models and areas from photos - 3DtoMe project.

Production-grade execution: privacy-by-design and GDPR alignment, human-in-the-loop where needed, model evaluation, telemetry and monitoring, CI/CD and MLOps.

Stack: Azure OpenAI Service, Azure AI Vision and Document Intelligence, Azure Speech, Azure ML, Ultralytics YOLOv8, OpenCV, Python, Swift and SwiftUI. When useful, we run on-device with Core ML and plug AI into Spatial Computing experiences (AR/VR) on Apple Vision Pro.

Explore our work in Spatial Computing, see 3DtoMe and Spatial Plan, meet the team or contact us to scope your MVP or rollout.

References: Microsoft Azure AI (accessed 2025); Ultralytics YOLOv8 Docs (accessed 2025); Apple visionOS (accessed 2025); n8n Docs (accessed 2025).