About Univault Technologies
We build the tool and the network for the health economy.
Univault Technologies LLC is a biosignal AI research company based in Salt Lake City, Utah, founded in 2024 by Philip Phuong Tran (CEO) and Anh T Do, PhD (Research Director). The company builds the General Learning Encoder (GLE), a subject-invariant biosignal foundation model that achieved over 13x more improvement than the next best team at the NeurIPS 2025 EEG Foundation Model Challenge. Univault also operates ParagonDAO, the verification network for health model governance. Patent pending (US Provisional 63/985,936).
Why We Exist
Every health researcher has data that could help patients — but no way to turn it into a product without building an entire company. Every patient has health signals their phone could read — but no models trained to interpret them.
Univault exists to close that gap. We built the universal encoder that turns any health signal into math, and the governance network that ensures the math is accurate before patients depend on it.
Two Products. One Architecture.
BAGLE
The Lab Equipment
Brain-AI General Learning Encoder. The universal health signal encoder powered by DCT-II frequency-domain processing. Any signal in — breathing, cardiac, EEG, voice, molecular sensors — 128 coefficients out. Builders train classifiers on those coefficients. The encoder does the hard math. 6 models published. API opens April 2026.
ParagonDAO
The Peer Review Board
The governance network that validates health models before patients rely on them. Certifies builders. Ensures quality. Manages the security layer (HF-Auth continuous authentication). 10% of all network fees fund one mission: preventing loss of life. Published whitepapers, open standard.
Why both? Because the tool without governance is dangerous — unvalidated health models loose in the world. And the governance without the tool is just a committee with nothing to govern.
The Builder Economy
Not One Company with 1,000 Employees
Thousands of founders building on one protocol. Each brings domain expertise in their health vertical. The encoder handles the math. ParagonDAO validates the quality.
The Mission Fund
10% of all network fees fund crisis detection, community health screening, and free GLE access for crisis organizations. 988 suicide prevention. Community health workers. The reason the network exists.
The Team
A husband-and-wife research team building the infrastructure for population-scale health AI.
Philip Phuong Tran
Founder & CEO
Architect of the GLE (General Learning Encoder) framework and the subject invariance methodology. Named inventor on the GLE patent application (rights assigned to Univault Technologies LLC). Based in Salt Lake City, Utah.
Applied Mathematics & Signal Processing
Master's in Mathematics and Statistics. Former lecturer at California State University. GLE's DCT-II frequency-domain architecture is grounded in this mathematical foundation.
Anh T Do, PhD
Co-founder & Research Director
PhD researcher. Lead author of Tín Hiệu và Hệ Thống (Signals and Systems), a signal and control theory textbook for Hanoi University of Science and Technology (HUST), Vietnam's leading engineering institution. Based in Salt Lake City, Utah.
Signal & Control Theory
Her academic foundation in signal processing directly underpins GLE's convex optimization methods for subject-invariant biosignal encoding.
Research Foundation
Everything we build is grounded in published research and verified benchmarks.
GLE Encoder
Over 13× improvement
over next best team at NeurIPS 2025 EEG Foundation Model Challenge — over a thousand competing teams
Breathing Biometrics
96.8% accuracy
identification across 97 participants using nasal airflow patterns alone
Patent Portfolio
2 provisionals filed
GLE universal encoder + piezoelectric textile biosignal system with continuous authentication