Computer vision demo
SignScribe
Emergency accessibility prototype for limited sign-language phrase recognition, translation, and spoken output.
Projects
Supporting work across CV, speech, validation, and interfaces.
These projects show the foundations behind the flagship case studies: computer vision, real-time inference, multimodal interfaces, and deterministic-first validation systems.
Production ML systems
Real-time WebRTC Vision
Object detection and OCR-style inference over live WebRTC streams, including sub-100ms frame inference in LinkedIn evidence.
Multimodal interface
Video-driven Avatar
Facial-expression mapping into 52 ARKit blendshapes for real-time 3D avatar animation in browser-based communication flows.
Validation platform
RuleHub AI
Rules-as-data validation platform concept using deterministic checks first and selective AI judgment for ambiguous cases.