Ambient multimodal AI for mental health
Mental health, measured as clearly as a scan.
Deliberate AI reads facial expression, voice, speech, movement, eye behavior and vital signs from an ordinary video visit, then turns them into objective measures for clinical trials, clinics and telehealth.
- First AI/ML initiative accepted into the FDA ISTAND Pilot Program
- Built on the 2009 study that started the field
- 600+ papers by our team
7 signals captured
Signal quality: high§ 01
Every other field of medicine has a scan.
Mental health still has a questionnaire.
Self-report drifts. Raters disagree. Symptoms go unseen between visits. The cost shows up everywhere.
Rater variability and subjective endpoints shrink effect sizes and inflate sample sizes.
Measurement-based care stalls when every measure is another form.
Visits happen at scale, but the signal inside them goes unmeasured.
§ 02
How it works
Hundreds of signals. One clear measure.
From a standard video recording, our engine extracts hundreds of behavioral and physiological features, fuses them, and returns interpretable measures that clinicians and researchers can trust.
- Facial expressionsaction units, expressivity and affect dynamics
- Vocal acousticsprosody, pitch, energy and timing
- Speech behaviorspauses, rate and turn-taking
- Head & body motionmovement dynamics and synchrony
- Oculometry & pupillometrygaze, blinks and pupil response
- Vital signsheart rate and breathing, read from video
- Computational linguisticswhat is said, and how it is structured
FIG. 01Seven modalities, one measurement engine. Features are extracted from a standard video recording.
01
Capture
An ordinary video visit or recording. No special hardware.
02
Extract
Hundreds of behavioral and physiological features.
03
Fuse and measure
Multimodal models return explainable, clinically meaningful measures.
§ 03
One engine. Four ways to put it to work.
Choose the path that fits your work. Every product runs on the same measurement core.
§ 04
The science
We didn't join this field. We started it.
In 2009, our co-founder and Chief Scientist Jeffrey Cohn, PhD, and colleagues published the first study to detect depression from facial actions and vocal prosody. Deliberate AI turns that legacy into tools built for real clinical work.
- 2009
- The first multimodal psychopathology study
- 600+
- Papers published by our team
- 64k
- Citations
- First
- AI/ML initiative accepted into FDA ISTAND
- 2009Landmark multimodal depression study
- 2020Deliberate AI founded
- 2024AICOA® accepted into FDA ISTAND
- TodaySEQUOIA clinical validation program
- 2009Landmark multimodal depression study
- 2020Deliberate AI founded
- 2024AICOA® accepted into FDA ISTAND
- TodaySEQUOIA clinical validation program
FIG. 03From the first multimodal study to regulatory qualification.
§ 05
The instrument, not the oracle.
Our models measure. Clinicians decide. Every result is explainable, every workflow keeps a human in the loop, and every recording is handled with the care mental health data demands.
Human in the loop
Designed to augment clinical judgment, never replace it.
Explainable by design
Interpretable models show which signals drove each result.
A regulatory path, not a promise
Pursuing FDA qualification of AICOA® through the ISTAND Pilot Program.
Privacy and security
§ 06
Clinicians and scientists, at the same table.
Our team blends clinical acuity with machine learning and careful software design, shaped by firsthand experience of where mental health care falls short.
- Marc AafjesFounder & CEO
- Jeffrey Cohn, PhDCo-founder & Chief Scientist
- Marcelo Cicconet, PhDData Science Director
- Katie Aafjes-van Doorn, DClinPsy, MScAssessment Director
FIG. 04Leadership across clinical science, assessment and data science.
See what your visits have been saying.
Tell us where you work and we'll show you the product built for it.
Request a DemoWe'll route your request to the right team.