MedTech · Digital Health · AI Safety

Safer control for future cortical visual prostheses.

AVCR is a history-, geometry-, intervention-, and uncertainty-aware software control layer designed to decide when to act, when to abstain, and what evidence must be preserved before stimulation.

Explore AVCR View pitch deck
Pre-clinical, non-human-use prototype. No clinical efficacy or patient-use claim.
Closed-loop AVCR architecture linking scene, gaze, history, geometry, safety policy and future cortical stimulation
Closed-loop artificial vision architecture: from scene to decision-certified control.
01History-aware state
02Geometry-aware mapping
03Uncertainty & abstention
04Auditable HIL evidence

The problem

Cortical vision cannot be treated as pixel writing.

Useful artificial vision is a closed-loop control problem. Perception can depend on the person, gaze, recent stimulation, cortical geometry, device state, and task. A static camera-to-stimulation map can erase history that changes the next percept.

Reconstruct

Build a task-specific state

Link scene, gaze, device state, stimulation history, feedback, and relevant context.

Map

Respect cortical geometry

Preserve local structure and support remapping when gaze, maps, or interfaces change.

Certify

Bound the decision

Use uncertainty and response-diameter gates to determine act, limit, wait, or abstain.

Verify

Produce replayable evidence

Connect software decisions to hardware-in-the-loop logs, provenance, and safety tests.

AVCR geometry-aware path from scene structure to cortical stimulation geometry
From retinal scene structure to geometry-constrained cortical stimulation planning.

Current stage

P0-B: a safe bench bridge from theory to translational evidence.

AVCR’s current physical prototype is designed for non-human, non-stimulating hardware-in-the-loop verification. It does not include an electrode connector, implantable current source, patient connection, or clinical-use authority.

  • E-stop and watchdog verification
  • Power-recovery and fail-closed behavior
  • Eight-channel timing and fault injection
  • Immutable logs, firmware hashes, and evidence packets
Read the full Hub71 pitch deck
P0-B safe bench prototype architecture
P0 proof-of-concept results summary
Evidence dossier montage

Abu Dhabi roadmap

A focused 90-day translational programme.

Anahita intends to establish Abu Dhabi as AVCR’s MENA engineering, validation, and commercialization base, with one founder relocating long-term if selected by Hub71+ Life Sciences.

01

Month 1

Establish

Set up the operating structure, relocate a founder, complete the IP and regulatory data room, and map clinical, regulatory, investment, and device partners.

02

Month 2

Verify

Complete independently witnessed HIL testing for timing, watchdog, E-stop, power recovery, fault injection, logging, cybersecurity scope, and abstention behavior.

03

Month 3

Partner

Secure a scientific or clinical collaboration pathway, finalize UAE and FDA regulatory roadmaps, recruit priority advisors, and prepare the next financing round.

About Anahita

Multidisciplinary science translated into safety-critical engineering.

Anahita is developing AVCR at the intersection of neurotechnology, control theory, medical-device software, evidence architecture, and patient-specific adaptation.

The programme is led by Esmaeil Farshi, founder, inventor, and author of the AVCR scientific framework.

AVCR

Decision-certified control for a safer artificial-vision future.

Start a conversation