Remember.
Always.
ZAKR is a closed-loop neural memory augmentation platform - a lightweight wearable and on-device AI system designed to help reinforce the neural pathways behind memory encoding and recall, in real time, under physician supervision. Built first for people living with Alzheimer's disease and mild cognitive impairment, and for the families who walk alongside them.
One closed loop, five quiet steps
ZAKR follows a continuous five-stage process to sense, understand, store and - when it's genuinely needed - gently help bring a memory back.
Acquire
ZAKR listens and senses gently in the background - noticing conversations, places, faces and routines without ever interrupting the moment.
Encode
An on-device model decides what's meaningful and worth holding on to, turning it into a secure memory record in a fraction of a second.
Store
Each memory record is sealed with hardware-grade encryption and saved safely on the device itself - nothing is lost, even if it's switched off.
Detect
ZAKR continuously compares what it senses against a baseline built just for that person, watching quietly for moments of difficulty.
Replay
When it's appropriate - and only within carefully governed limits - ZAKR offers a gentle nudge: a familiar voice, sound or cue to help the memory surface.
Designed around people, not just protocols
Keeps Memories Alive
Helps recall the moments that bring joy, identity and meaning back into focus.
Strengthens Connections
Supports easier conversations, closer bonds, and less day-to-day frustration for families.
Empowers Independence
Helps people do more on their own, with a quiet sense of confidence.
Supports Brain Health
Continuous, passive monitoring helps keep the mind engaged over time.
Built With Trust
Every memory record stays encrypted, on-device, and is never shared without consent.
The NV-Band, at a glance
A single page from our printed brochure - the same five-stage loop and everyday benefits described above, designed for families meeting ZAKR for the first time.
- Works quietly in the background, 24 hours a day, with care and privacy.
- Long battery life - charges once every few days.
- Private and secure. Data stays the user's own, always protected.
- Lightweight and comfortable, made for all-day wear.
Built by people who set out to solve this
ZAKR was founded to close one of the most painful gaps in medicine today - the steady loss of memory and identity that comes with Alzheimer's and related conditions.
Dr. Hazeem
With over 25 years of experience across global technology enterprises, Dr. Hazeem is the sole inventor behind ZAKR's full portfolio of innovations - now working to bring a neural memory platform from concept to clinic.
Muhammed Aamir
Co-founder of ZAKR, helping shape the platform's direction as it moves from research and intellectual-property development toward clinical validation and global partnerships.
A platform shaped around dignity, safety and memory
From the underlying neuroscience to the hardware safety architecture and the AI that governs it - every layer of ZAKR is designed with one outcome in mind: helping someone hold on to who they are.
32+ innovations. Three patent families. One closed loop.
Every part of ZAKR - the algorithms, the wearable hardware, and the AI that governs it - is independently developed, documented and protected as its own body of work, while operating together as a single closed-loop system.
AI Governance & Protocol Safety
A language-model layer that proposes stimulation protocols - but never has the final word. Every proposal, regardless of source, is checked against the same hardware-enforced safety limits before anything is delivered.
Neural Memory Packet System
The intelligence layer - the algorithms, cryptographic protocols and data architecture that decide what gets remembered, how it's protected, and when it's safe to bring back.
NV-Band Wearable Platform
The wearable itself - a bilateral, lightweight headset built around a hardware safety stack that no software update, voice command, or unexpected input can override.
The innovations that make the loop possible
A closer look at the systems working together beneath the surface - each one independently documented, and each one essential to the platform as a whole.
NMRGS
A three-gate governance system - integrity verification, frequency limits, and minimum intervals - that prevents memory replay from happening too often, while never limiting what the person themselves chooses to revisit.
AES v3 - Four-Tier Vault
A progressive authentication system designed to remain accessible across the full course of a progressive condition - so a vault is never permanently locked, even as biometric signatures naturally drift over time.
Neural Memory Packet
A structured, encrypted record created for every memory event - capturing what was sensed, how confident the system was, and exactly how it may (and may not) be replayed.
EPT-00 Emergency Mode
A dedicated zero-stimulation emergency mode. Voice-activated alerts reach a caregiver instantly - without ever bypassing the hardware limits that keep stimulation safe.
HBCI & QAW Coherence
Composite indices combining cardiac, neural and vascular signals into a single coherence score - used to recognise the moments most worth remembering.
Cognitive Longevity Index
A longitudinal cognitive health score, built passively from everyday signals - with no test to take, no task to complete, and no clinic visit required.
Cross-Modal Replay
Sound, sight and scent cues synchronised within milliseconds of a stimulation pulse - drawing on published research linking multi-sensory cues to stronger recall.
Federated Learning
The model improves over time using on-device learning and strong differential-privacy guarantees - with consent-gated participation and no raw neural data ever transmitted.
Seven-Layer Safety Stack
From a physical kill switch to a passive hardware current limiter, each layer operates independently - so no single software issue can ever compromise safety.
taVNS & cPWV
Gentle auricular nerve stimulation alongside pulse-wave-based vascular monitoring - derived from sensors already on the device, with no additional hardware required.
Bone-Conduction Isolation
A three-layer isolation design keeps audio cues delivered through bone conduction from ever contaminating the EEG signals being recorded at the same time.
Source-Agnostic Enforcement
Whether a stimulation protocol comes from a voice command, a companion app, a clinician, or an emergency trigger - it passes through exactly the same hardware safety checks, with no exceptions.
Thirty future directions, already mapped
Beyond the platform shipping today, ZAKR has charted thirty future innovation directions - extensions of the same closed-loop foundation into new conditions, new hardware forms, and new research infrastructure.
Neurotechnology
Biometric Security
AI Governance
Clinical Applications
Research Infrastructure
Wearable Hardware
Three layers. One closed loop. Under 250 milliseconds.
ZAKR is built as three independent layers working in concert - AI governance on top, the intelligence layer in the middle, and purpose-built hardware underneath. Each layer can be understood, tested and certified on its own, while the platform behaves as a single, fast, closed loop.
Protocol generation, with hardware having the final word
A fine-tuned language model proposes stimulation protocols from voice commands, a companion app, or clinical input. Every proposal - without exception - is checked against the same hardware-enforced limits before it can ever reach the wearable.
Encoding, governing and protecting every memory
On-device inference turns raw signals into structured memory records, governs how and when they may be replayed, and keeps a four-tier biometric vault accessible for life - all without sending anything off the device.
The NV-Band - sensing, stimulation and safety in one wearable
A bilateral, lightweight headset combining multi-channel neural sensing, gentle stimulation, and a seven-layer hardware safety stack that operates entirely independently of any software.
From sensing to a gentle cue - in under 250 milliseconds
Every stage of the loop happens on-device, in real time. Detection runs continuously in the background; the entire response - from the moment a need is recognised to the moment a cue is delivered - completes in well under a quarter of a second.
Acquire
Multi-channel neural sensing and ambient audio capture, with on-device noise rejection.
Encode
On-device inference structures the moment into an encrypted memory record.
Store
The record is sealed with hardware-grade encryption inside a certified secure element.
Detect
Ongoing comparison against a personalised baseline, watching for moments of difficulty.
Replay
A governed, multi-sensory cue is delivered - only once every safety check has passed.
A wearable built for all-day, every-day use
Comfort and safety were treated as first-class requirements from the start - because a device that helps with memory only works if someone is willing to wear it.
Seven independent layers. No single point of failure.
Each layer below operates on its own - most without any reliance on software at all. A failure, bug, or unexpected input at any one layer cannot defeat the layers around it.
Physical kill switch
A hardware switch that immediately and irreversibly halts all stimulation, independent of any software state.
Sustained voice gate
Voice commands are only accepted after a deliberate, sustained activation phrase - reducing accidental or adversarial triggers.
Integrity verification
The device cryptographically verifies its own safety configuration against a trusted reference before acting on any instruction.
Protocol constraints
Every proposed protocol - type, waveform, placement and ramp rate - is checked against an approved reference library.
Safety envelope limits
Absolute, hardware-enforced limits on current, frequency, session length and sessions per day - applied identically to every source.
Hardware current limiter
A passive component physically caps stimulation current, regardless of any other failure in the system.
Galvanic isolation
Stimulation and sensing circuitry remain electrically isolated from one another at all times.
A vault that stays accessible - for life
Memory naturally changes how a brain's signals look over time. ZAKR's authentication system is designed around that reality, with four progressive tiers that ensure no one is ever permanently locked out of their own device.
Primary brainprint
A personal neural signature, established over multiple sessions and refreshed over time as the baseline naturally evolves.
Backup brainprints
Additional signatures captured at intervals, each independently capable of unlocking the vault on its own.
Voiceprint
A secondary biometric fallback, used only if neural-signature matching is temporarily unavailable.
Guaranteed access
A clinician-assisted override that always grants access to essential medical records - regardless of how much biometric drift has occurred.
Built on published neuroscience, not novelty for its own sake
Every subsystem inside ZAKR is grounded in a specific body of published research - from how memories are encoded and reconsolidated, to how multi-sensory cues and heart-brain coherence influence recall. The platform's contribution is bringing these findings together into a single, real-time, on-device loop.
Phase-Amplitude Coupling & Personal Baselines
Continuous monitoring of phase-amplitude coupling - a well-studied marker of memory encoding - compared against a slowly-updating personal baseline, rather than a fixed population threshold.
Memory Reconsolidation
Each time a memory is recalled, it briefly becomes changeable again. ZAKR's replay governance reflects this - limiting how often and how soon a memory is reinforced.
Multi-Sensory Cueing
Combining sound, sight and scent around a single moment of recall draws on research showing multi-sensory cues can meaningfully improve memory retrieval.
Heart-Brain Coherence
Cardiac, neural and vascular signals are combined into a single coherence measure - built on a longstanding research tradition linking heart rhythms and brain state.
Passive Cognitive Monitoring
A longitudinal cognitive health score, computed from everyday physiological signals - without requiring a cognitive test, a clinic visit, or active effort from the user.
Slow-Wave Sleep Support
Gentle, phase-locked stimulation during deep sleep, informed by research on slow-oscillation stimulation and its role in overnight memory consolidation.
The literature ZAKR builds on
A selection of the published research underpinning ZAKR's approach - spanning memory science, neuromodulation, cardiac coherence, and closed-loop neural interfaces.
A three-paper research series
ZAKR's research program is structured as three sequential publications, each building the clinical and scientific case for the platform's core systems.
Real-world PAC monitoring as an Alzheimer's biomarker
Establishing continuous, non-invasive phase-amplitude coupling monitoring as a real-world biomarker - a foundation for future drug-trial endpoint use.
Security and usability of a progressive biometric vault
Evaluating the four-tier authentication system's usability and security across populations with progressive neurological conditions.
Closed-loop stimulation with governed replay - safety and early efficacy
Reporting safety outcomes and preliminary efficacy signals from the full closed-loop system, including its replay governance design.
Working with researchers worldwide
ZAKR's research program is designed for collaboration with academic and clinical partners across multiple regions - sharing access to its research platform, co-authoring publications, and contributing to a privacy-preserving body of longitudinal data.
Built on principles, not just protocols
ZAKR is being developed for a population that depends on trust - patients living with cognitive change, and the families and clinicians supporting them. These six principles guide every design and business decision, and sit alongside the formal safety architecture described elsewhere on this site.
Data Sovereignty
Raw neural data, biometric keys, and memory vault contents never leave the physical device. Privacy is a property of the architecture, not a policy layered on top.
Patient Dignity & Cognitive Identity
The four-tier vault and its guaranteed-access tier ensure identity-linked records remain reachable throughout the course of a progressive condition - supporting continuity of care and of self.
Non-Harm, Safety-First
Passive hardware safeguards - current limiters, safety envelope validation, galvanic isolation, and a zero-stimulation emergency mode - cannot be overridden by software, a voice command, or a system failure.
Equitable Access
An on-device AI architecture removes the need for constant connectivity, supporting use in both well-resourced clinical settings and lower-resource environments alike.
Transparent AI
A public model card documents measured performance, safety testing, and deployment architecture before any commercial release - open to review by clinicians, regulators and researchers.
Outcome-Aligned Profit Sharing
ZAKR's commercial model is structured around equity-participation and revenue-sharing rather than fixed-return debt - aligning returns with real clinical deployment outcomes over time.
Prescription-supervised, by design
ZAKR is being developed as a medical device. That distinction shapes how the platform is tested, documented, and ultimately brought to clinics.
Therapeutic device pathway
The closed-loop stimulation platform is being developed toward a premarket-approval-class regulatory pathway, reflecting its role as an active therapeutic device used under physician prescription.
Monitoring-only pathway
A monitoring-only configuration - without stimulation hardware - is being developed toward a lower-risk regulatory classification, suited to healthy-aging and wellness monitoring under clinical guidance.
AI/ML medical device framework
The AI governance layer is designed around predetermined change-control principles for AI/ML-based medical devices, with every model update re-tested and re-documented before release.
Commitments that apply from day one
No data ever sold
Neural data, biometric keys and memory records belong to the person wearing the device - never sold, and never used for advertising.
No permanent lockouts
The four-tier authentication system is designed so that no one is ever permanently denied access to their own medical identity records.
No silent AI changes
Every update to the on-device AI model is documented in the public model card before it is deployed - including its measured safety performance.
No hidden commercial terms
ZAKR's commercial structure is built on transparent, outcome-aligned profit sharing - not interest-bearing debt or hidden minimums.
AI transparency, published before deployment
ZAKR's on-device AI proposes stimulation protocols - it never delivers them directly. This page documents how that model is built, tested, and governed, in line with international frameworks for AI-based medical devices. It is republished, in full, before every model update.
ZAKR Protocol Generation Model - Card
Published prior to deploymentBase model & fine-tuning
A language model fine-tuned on a curated dataset of at least 500 peer-reviewed transcranial stimulation publications, specialising it for the language and parameter ranges relevant to neuromodulation protocols.
Constitutional alignment + clinician preference
The model is aligned using a constitutional-AI approach combined with reinforcement learning from physician preference feedback - favouring conservative, literature-supported parameter choices over novel or extreme ones.
Hallucination rate
Maximum permitted hallucination rate of 0.5%, measured on a held-out evaluation set of at least 500 clinical prompts. Every output is independently annotated by at least three board-certified neurologists, with strong inter-rater agreement (Cohen's κ ≥ 0.80). A "hallucination" is defined as any proposed parameter outside hardware safety limits, outside published literature ranges, or specifying an unapproved stimulation configuration.
Eight-category attack testing
Tested against at least 2,000 adversarial prompts and 500 positive controls across eight categories, with a pass threshold of less than 1% breach rate and at least 95% positive-control acceptance.
Three ways to run the model - one safety standard
The model can run in any of three configurations. All three submit every generated protocol to the identical hardware safety validation described in Technology.
Companion device
Runs on a paired smartphone or tablet. Seconds-level latency, with an offline preset cache.
Local edge server
Runs on a local network server. Faster response, with a larger offline preset library.
On-device micro-model
Runs directly on the wearable. Sub-200ms latency, fully capable offline.
Always a safe path forward
If the model is unavailable, or its output does not pass validation, the device falls back to clinician-configured preset protocols - validated through the same hardware safety stack as any AI-generated proposal.
Every update, re-tested and re-published
Before any model update is deployed - including updates produced through on-device federated learning - the hallucination rate and adversarial test suite are re-run in full, and this model card is republished with the new results.
AI Safety Advisory Board co-signature
Each published version of this model card is reviewed and co-signed by ZAKR's AI Safety Advisory Board prior to deployment.
Let's talk
Whether you're a clinician, a researcher, a potential partner, or simply curious about ZAKR - we'd like to hear from you. The founding team reads every message personally.
Research & Academic Partnerships
Open to collaborations with academic and clinical research groups worldwide - including data-sharing agreements, co-authored publications, and joint grant applications.
Clinical & Investigator Interest
If you're a clinician or investigator interested in the platform ahead of regulatory clearance, please get in touch - we're building our early research network now.
Partnerships & Investment
For commercial partnerships, licensing discussions, or investment inquiries, please use the form and select "Partnerships & Investment."