9
Designing repeatable, explainable architectures that bridge cognitive theory, linguistics, and hard data pipelines. Large Language Models are utilized as tools to think, reduce known unknowns, and accelerate ideas by providing expertise on demand.
We prioritize structural truth over black-box outputs, relying on classical measurement systems and custom extraction pipelines to establish ground truth.
A feature extraction and macroprosody mapping initiative analyzing vocal telemetry and voice physics to establish ground-truth data across 150-200 languages.
Status: 7-language sample set active, 100 languages processed, ~1.5m samples assessed
Data Samples on HF AboutLeveraging progress and discoveried from the Human Prosody Project, we have begin feature and physics extraction on several Deep Fake collections.
Status: 500k Deepfakes assessed.
System DetailsUtilising our cllection of real and synthetic voice data we are able to develop a light weight classifier to detect audio deepfakes based on their inability to properly emulate biology
Status: Model development active.
System DetailsAn emotive Text To Speech model designed to utilize structured prosody and vocal physics for programmatically human, emotive vocal outputs.
Status: Model development active.
System DetailsThe Digital Dungeon Master. A logic-driven, solo campaign engine built on the 5e SRD ruleset with some adapted mechanical elements.
Status: Engine development active.
Launch DDMUniversal Cognitive-Memory Architecture. A model-agnostic neurosemantic OS design rooted in cognitive linguistics, neuroscience, and educational theory.
Status: Architecture drafting.
View ArchitectureA centralized knowledge base and retrieval system structuring educational materials, statistical information, and internal business data.
Status: Database structuring.
Access Database