Aarakshak - Evidence Workbench for Verified Presence

Evidence workbench

Build, run, evaluate, and govern every verification decision.

Aarakshak is not a generic face API page. It is an operating surface for teams that need to prove who was present, what the model decided, which evidence was reviewed, and why the final action was allowed.

Build

Configure endpoint contracts, tenant keys, thresholds, consent fields, and review bands before the first production capture.

Run

Operate compare, verify, gallery search, attendance, and secondary-camera screen checks from one consistent evidence pipeline.

Evaluate

Inspect match quality, liveness signals, screen-risk overlays, false-reject cases, and held-out promotion evidence before changing thresholds.

Govern

Keep signed packets, reviewer notes, retention rules, model versions, and audit exports connected to the decision they support.

Core journeys

The four journeys buyers actually need to ship.

The platform is organized around operational journeys, not a scattered product catalogue. Each path has the same trust model: clear inputs, calibrated thresholds, reviewable evidence, and an accountable final decision.

Assessment integrity Assessfy-ready exam evidence

Identity plus screen evidence

Verify the candidate, monitor secondary-camera screen evidence, route risky moments to human review, and export a defensible packet for disputes or audit.

Candidate verify Screen check Review queue Signed packet

Attendance and campus operations mUniCampus-ready presence records

Classroom, lab, office, event

Enroll once with consent, verify at check-in, and support UGC-aligned audit records without turning daily attendance into a surveillance dashboard.

UGC-aligned Audit export Mobile SDK

KYC and account recovery 1:1 evidence for regulated builders

Compare, quality, review

Compare a document image or enrolled reference against a live capture, include quality signals, and send borderline decisions to reviewer-assisted recovery.

Face compare Quality gates Reviewer notes

Gallery search and access Tenant-scoped identity lookup

Search, verify, decide

Search a bounded gallery, verify the top candidate, and keep the grant or deny decision tied to the exact model result and reviewer action.

1:N search Tenant boundary Access audit

Sovereign API integration Contracts, keys, logs, and guardrails

Developer-ready, governance-first

Give builders the API explorer speed of Google AI Studio with the usage ledger, event traceability, and policy controls expected from enterprise AI platforms.

API explorer Scoped keys Usage ledger

Platform architecture

A workbench model inspired by the best AI developer platforms.

Stripe-like debugging, Linear-like workflow clarity, Foundry-style evaluation gates, and Google AI Studio speed only matter if they are grounded in institutional controls. Aarakshak brings those patterns to verified-presence operations.

API layer Stable contracts for product teams

Explorer, snippets, scoped keys

Compare, verify, search, enroll, and screen-check endpoints expose stable contracts with examples, tenant-scoped keys, and usage traces.

API Explorer Endpoint contracts Scoped keys Usage ledger

Evidence layer Decision packets that survive review

Scores, media, thresholds, model version

Every decision binds source media, scores, thresholds, timestamps, tenant, reviewer action, and model version into an exportable evidence packet.

Signed packet Threshold record Review notes Exportable audit

Review layer Human gates for borderline outcomes

No hidden auto-disqualification

Borderline identity, liveness, and screen-risk cases move into a reviewer queue with context, notes, and final disposition captured beside the model result.

Review band Reviewer ID Appeal-ready

Operations layer Admin control plane for deployment teams

Keys, retention, promotion evidence

Admins manage tenant keys, usage, model promotion evidence, retention posture, and customer scenarios without leaving the product surface.

Retention rules Promotion gate Scenario loader

Operating modes

From pilot to production without changing the trust model.

Every deployment starts narrow, then expands through the same governed path: controlled API use, reviewer calibration, model evaluation, and production evidence export.

Attendance system

enroll → verify → usage

Enroll the employee or student once. Verify a live capture at check-in. The fastest workflow to ship and the easiest to run on existing classroom or office hardware.

Best for classrooms, offices, events
Watch false rejects under poor lighting — tune threshold per location

Live proctoring

enroll → verify → liveness

Enroll an ID or reference image at session start. Verify periodic webcam captures during the exam. Review low-score events with a human invigilator post-hoc — never auto-disqualify a candidate.

Best for exam identity continuity
Watch set a review band near threshold; keep auto-pass and auto-fail strict

KYC: ID vs camera

compare → quality → review

Upload an ID document photo. Capture a live camera shot in the same session. Run a 1:1 face compare with quality grading on both inputs. Bind the decision to an audit trail.

Best for fintech onboarding, account recovery
Watch add anti-spoof before any production rollout — this flow alone is not anti-spoof

Security access

search → verify → audit

Search a tenant-scoped gallery against a live capture. Verify the top candidate before granting access. Every grant or deny is timestamped, signed, and audit-logged.

Best for ministry passes, restricted floors, BFSI back-office
Watch use stricter thresholds for high-risk areas; pair with a second factor at the door

Governance

High-risk verification needs controls, not just confidence scores.

Aarakshak keeps sovereignty concrete: India-hosted or on-prem deployment, explicit human review gates, evaluation evidence before promotion, and audit packets buyers can actually inspect.

1

Deployment boundary is explicit

Run India-hosted or on-prem with tenant-scoped storage, known sub-processors, and a clear data-retention posture before pilot launch.

2

Human review is part of the product

Borderline decisions carry reviewer identity, notes, timestamps, and final disposition so operators can defend outcomes without guesswork.

3

Evaluation gates before promotion

Model and threshold changes require held-out performance evidence, workflow context, and operator sign-off before production use.

4

Audit packets are first-class

Exports connect the capture, score, threshold, reviewer action, tenant, and model version into one decision record for internal or external review.

Best fit

Built for buyers who need evidence, controls, and speed together.

Assessment platforms

Hiring, certification, and exam teams that need identity continuity plus screen evidence.

Universities & institutions

Campus teams running attendance, exam, lab, and gallery workflows under real audit pressure.

Regulated builders

Fintech, hiring, and KYC products that need governed identity APIs without foreign-region dependency.

Secure operations

Public-sector and high-control environments that need on-prem capable deployment and strict review bands.

Sovereign verified presence
for India's institutions.

Assessfy

mUniCampus

Photo Finder AI