Built on published science — open to the people who make it.
Recalla is an early-stage research effort in cognitive and voice biomarkers. We're actively looking to work with researchers, clinicians, and institutions who share our commitment to doing this carefully. Here's how a collaboration can take shape.
Academic research
Joint studies, method development, and shared analysis of voice and cognition — with honest, co-authored outputs.
Clinical & hospital
Pilots, clinician feedback, and real-world evaluation in memory clinics, neurology, and eldercare settings.
Advisory roles
Clinical, scientific, and technical advisors who keep the work grounded and the claims honest.
Validation & benchmarking
Evaluating methods against established public benchmarks, and — as we generate data under consent — rigorous validation.
Design partners
Care providers and clinicians who help pressure-test the workflow with real needs, not assumptions.
Students & theses
Scoped projects for graduate researchers interested in speech, cognition, or responsible health AI.
The questions we most want to work on.
These are open, unglamorous, important problems — and the areas where the right collaborator would have the most impact. They map directly to our Evidence Hub.
- Within-person change over time. Moving beyond cross-sectional group classification toward validated longitudinal, personal-baseline signals.
- Task and prompt design. Check-in tasks with known structure that yield repeatable, interpretable measurements across sessions.
- Explainability clinicians actually use. Per-signal contributions that support human judgement rather than replace it.
- Robustness in the real world. Performance across languages, accents, devices, and everyday recording conditions.
- Adherence and usability. What makes people — and their care partners — willing to check in consistently over months.
How we handle sensitive data — before we ask you to trust us with it.
Voice can carry health-related signal, so we treat it accordingly. These principles govern any collaboration involving real participant data.
Governance in practice
- Consent-first. Any voice or health-related data is processed only under explicit, revocable consent, for a stated purpose.
- Special-category handling. Heightened protection, data minimisation, and least-privilege access controls.
- Ethics review. Studies are designed for appropriate ethics/IRB oversight, with participant welfare first.
- Purpose limitation. Data is used only for the agreed research purpose — never ad profiling, never sold.
- Synthetic by default. Public demos use synthetic data; live features run on-device and store nothing.
Transparency
We publish our limitations, not just our hopes — the Evidence Hub separates external science from our own unproven work.
Scientific integrity
We claim only what's true and defensible, and we frame our approach as a hypothesis under evaluation.
Human-in-the-loop
Recalla is non-diagnostic and designed to inform a professional's judgement — never to replace it.
Formal data-sharing, IP, and study terms are agreed per collaboration. See our Privacy Policy and Legal & notices for baseline commitments.
What we bring — and what we're looking for.
What we bring
- A working prototype and an explainable feature-and-scoring pipeline.
- An open technical framework and an interactive, synthetic-data demo.
- A founder who ships, and a transparent, verified evidence base.
- A genuinely consent-first, non-diagnostic posture.
What we're looking for
- Access to clinical populations and real-world settings.
- Validation and study-design expertise.
- Grant and funding partnerships for prospective work.
- Complementary datasets and domain guidance.
Working on cognition, speech, or responsible health AI?
Tell us about your group, your area of interest, and what you have in mind. Early, exploratory conversations are welcome — no formal proposal needed to say hello.
Contact us about collaborationHelpful to include: your institution or group, your role, the collaboration type above that fits best, and a sentence on what you're proposing.