The Intelligence
HALO transforms multimodal biophysiology into cognitive-state estimates through a disciplined pipeline — synchronized acquisition, robust preprocessing, engineered features, a tiered model stack, and per-user personalization — all expressed probabilistically with confidence and explanation rather than false certainty.
Robustness before precision
Biological signals are noisy and change with sleep, stress, and hydration. HALO is built to give a reliable estimate anyway.
Preprocessing the signal
First HALO cleans the data — removing drift, electrical hum, blinks, and movement before any analysis.
Features before models
Before any AI, HALO extracts meaningful measurements: rhythm strength, heart-rate variability, skin response, and more.
A tiered model stack
Simple, explainable models provide a reliable baseline; deeper models add nuance; personal models adapt to you.
States as probabilities, not labels
HALO never says simply 'focused or not.' It gives a probability and how confident it is.
The Harmonic Index — and why it's explainable
All signals combine into one intuitive score — and HALO always explains why the score moved.
A score is easy to show — but can it be proven?
Robustness and synchronized, quality-gated preprocessing come before any model.
Three tiers balance interpretability, power, and personalization — personalization is the moat.
Outputs are probabilistic, confidence-bounded, and explainable by design.