Where the
machine learns.
Most AI systems are built on problems that have already been solved. SignaVision works in areas where the problem is still open — where datasets are incomplete, models fail in real conditions, and meaningful progress requires original work.
This is not research for publication.
This is research that feeds live systems.
Focus Areas
Four active research directions.
01
Sign Language Recognition Models
Understanding sign language is not just visual detection — it is spatial, temporal, and contextual interpretation. The challenge is not accuracy in controlled environments. The challenge is generalization in the real world.
02
Multimodal AI Systems
Sign language does not exist in isolation. It intersects with text, speech, and context — yet most AI systems treat these as separate domains. This research explores how meaning can be shared across modalities, allowing systems to understand and translate between visual signing, language, and audio.
03
Accessibility Platform Architecture
Most accessibility tools are added after systems are built. This research focuses on what happens when accessibility is built into the system itself — at the protocol and data model level. LTI protocol design, video data structures, and real-time WebRTC pipelines treated as first-class concerns.
04
Knowledge Systems & RAG
Access to information is only useful if it can be understood and retrieved meaningfully. This research applies retrieval-augmented generation to accessibility contexts — allowing users to query large bodies of curated knowledge in natural language, grounded in verified data.
e-Forger Chamber Builder
An experimental system exploring how software can be generated — not written. The e-Forger Chamber Builder treats applications as structured intent, compiling them into working systems through composable node graphs. Instead of manually building interfaces and logic, the system constructs them from defined relationships between components. This is a meta-layer of software creation. It is not a normal project.
See it in the LabApproach
Research that ships.
There is no separation between research and engineering. Every model, system, and experiment is tested in real environments — with real users — and refined through direct feedback.
This loop is continuous:
Sign language is not a static dataset. It is a living, community-driven form of communication. Any system that attempts to understand it must evolve with it.
Collaborate on research.
We work with universities, accessibility organizations, and Deaf community programs to advance these systems in real contexts. If your work intersects with these areas, we are interested in building together.
Get in touch