BUILDING TODAY FOR TOMORROW’S NEEDS
The next wave of Smart: applied AI for language, skin health, customer experience, safer spaces and secure access. See what we are building — and where you can get involved.

Purpose: A natural, fluent experience for watching English video content in Farsi.
Current focus: Real-time translation, natural Persian speech generation, preserving the speaker’s tone, and keeping audio in sync with the video.
Next step: Build and evaluate a first prototype on short video content.
Audio is processed only to produce the dubbed stream; no speaker identification and no voice profiling.

Purpose: Simulating skin condition and showing possible care and treatment scenarios, to support more informed decisions.
Current focus: Modelling skin layers, analysing imaging data, and estimating likely changes over time.
Next step: Scientific validation of the base models with specialist data and clinician review.
Skin images are processed only with explicit consent. This project is a decision-support tool — not a replacement for a physician’s diagnosis or prescription.

The problem, the real need and the potential value are examined.
The core capability is built and tested at limited scale.
Quality, safety, privacy, fairness and reliability are assessed.
Tested in the real world with a defined scope, human oversight and clear success criteria.
The stage shown is each project’s current development status — not readiness for public release.
Behaviour analytics ≠ biometric security. Analytics is anonymous and aggregate — no facial recognition, no identity; biometric access control is a separate, consent-based system. No Smart project infers people’s emotions — aligned with the EU AI Act’s prohibited-practices rules.
Every project starts from a real need and must deliver a meaningful outcome for people.
Privacy, fairness, safety and human oversight are designed in from the start.
Progress and impact are assessed with clear criteria and measurable results.
Purpose: Responsibly studying the factors that shape first impressions in human interactions.
Current focus: Multimodal analysis of visual and behavioural cues, together with measuring and reducing model bias.
Next step: Internal evaluation across diverse scenarios and groups, before any limited pilot.
Anonymous analysis only — no identity recognition, no emotion inference, no biometric templates; aggregate insight, never individual profiles.

Purpose: Helping create safer work and public spaces by detecting movement patterns, crowding, dwell and unusual events — without attributing intent or judging individuals.
Current focus: Local (on-premises) camera processing, anonymous spatial-pattern analysis, and privacy-control design.
Next step: A privacy audit and a limited, supervised pilot.
No facial recognition, no identity, no emotion inference — aligned with the EU AI Act prohibitions; aggregate, privacy-preserving metrics by design, fully separate from any access-control system.

Purpose: Fast, secure access to authorised areas in workplaces, with privacy at the core of the design.
Current focus: Face liveness verification, local processing, permission management, and fallback paths for failure cases.
Next step: Hardware selection, security testing, and design of a controlled pilot.
Strictly separate from behaviour analytics: enrollment is explicit and consent-based, biometric data is used only for the access decision, and every decision is audit-logged.
