Advanced Engineering Lab

Physiological signal processing · EEG · BCI · Adaptive AI

Turning biological signals into systems you can inspect.

Advanced Engineering Lab develops physiological signal-processing and adaptive AI systems from wearable, neural, sleep, heart, activity, and recovery data — with uncertainty treated as part of the measurement.

Wearables and neural sensors only observe fragments of the body. AEL builds systems that determine what those signals can reliably support, quantify confidence, and then use AI to explain the result.

Current focusCollaborate
Animated occupancy map of a human silhouette with organ-layer activity

Approach

Deterministic estimates first. Language after.

Biological sensors provide noisy, incomplete observations. AEL computes an inspectable estimate, evaluates the quality of the evidence, attaches confidence, and only then uses language models to explain the result.

The model describes the estimate. It does not make it.

  1. //01

    Signal

    Incomplete biological observations.

  2. //02

    Estimate

    Deterministic, inspectable computation.

  3. //03

    Confidence

    Evidence quality made explicit.

  4. //04

    Interpretation

    Language explains what was measured.

Structure only where the record supports it

Every estimate carries its evidence.

AEL systems track which signals contributed to an estimate, how reliable they were, where signals disagreed, and what the available data cannot support.

Confidence is engineered into every output. Signal quality, missing data, degraded inputs, and disagreement determine the bounds of inference.

  1. //01

    Which signals contributed, and how reliable were they?

  2. //02

    How far does this reading sit from the person's own baseline?

  3. //03

    Did sensors agree, or did the evidence conflict?

  4. //04

    What does the available data not support?

One research problem. Two timescales.

Across days

CareTrace

CareTrace studies physiological change over time using heart, sleep, activity, recovery, and wearable data.

Across seconds

OMLAB

OMLAB studies neural and physiological state using EEG, HRV, voice, and multimodal signals.

Both ask the same question: how much can a system reliably infer from incomplete biological signals?

  1. Signal
  2. Estimate
  3. Confidence
  4. Interpretation

CARETRACE

//01

Personal baselines from Apple Health and wearables for sleep, recovery, heart, and activity — with signal confidence and optional family sharing.

OMLAB

//02

Neural and physiological research platform for short-timescale state from EEG, HRV, voice, and multimodal sensor streams.

Current focus

CareTrace

Understand your health patterns. Stay connected to the people you care about.

CareTrace turns Apple Health and wearable data into a personal baseline for sleep, recovery, heart, and activity. It helps people see what changed, how much evidence supports the estimate, and optionally share their health picture with family.

  1. Personal baseline
  2. Readiness
  3. Signal confidence
  4. Long-term patterns
  5. Family sharing

  • See change against your own baseline rather than only a population average.

  • Review readiness and long-term patterns with estimate confidence visible.

  • Connect sleep, recovery, heart, and activity in one health picture.

  • Share with family by choice; each person's baseline remains separate.

CareTrace iOS home showing resilience, sleep, and activity

Research

Core research areas

  1. //01

    Physiological signal processing

    Extracting usable information from noisy biological sensor streams.

  2. //02

    Personal baselines

    Understanding change relative to the individual rather than only population averages.

  3. //03

    Reliability-aware inference

    Making confidence, sensor quality, disagreement, and missing data part of the output.

  4. //04

    Multimodal fusion

    Combining signals only when the combined evidence justifies stronger inference.

  5. //05

    Neural interfaces

    Studying measurable relationships between neural signals, physiology, and state.

  6. //06

    AI interpretation

    Using language models to communicate computed findings rather than manufacture them.

What we do not claim

AEL defines each claim by the resolution and reliability of its source data.

  1. //01

    AEL systems are not substitutes for medical diagnosis.

  2. //02

    Consumer wearable signals are not equivalent to clinical measurements.

  3. //03

    EEG research does not imply thought decoding.

  4. //04

    Confidence depends on the available evidence and sensor quality.

  5. //05

    Language models do not generate the underlying physiological estimate.

We state what the signal can support — and stop there.

Collaboration

Work with AEL

We collaborate with research groups and technology teams working on wearable sensing, neuroscience, physiological AI, signal processing, and human-computer interfaces.

  • Physiological signal processing
  • EEG and BCI methods
  • Wearable sensor validation
  • Multimodal sensing
  • Personal baseline modeling
  • Uncertainty-aware estimation
  • Physiological AI systems
  • Research prototypes

Open to universities, neuroscience labs, wearable and health-technology companies, AI research organizations, sensor manufacturers, and strategic R&D partners.

Contact the lab

What may a system say about a body it can only partly observe?

That is the research question behind CareTrace and OMLAB: build from observable signals, make the evidence inspectable, and keep every interpretation within what the measurement can support.

Current focusCollaborate