Koraput District – Unified AI/ML Monitoring Platform (Extended Shareholder Report)

Executive Overview

Koraput is a geographically large, diverse district where outcomes in education, nutrition, health, and infrastructure are tightly coupled to last‑mile execution. Historically, monitoring has been department‑siloed and paper‑heavy: separate registers for attendance, mid‑day meals (MDM), inspections, grievances, and facilities—each with its own formats, timings, and escalation paths. The consequence is predictable: delayed visibility, inconsistent evidence, and weak accountability.

Xapny Solutions is deploying a Unified AI/ML Monitoring Platform that brings all critical public‑service signals onto a single, governed backbone—from classroom attendance and meal quality to facility uptime and grievance redressal. The platform is designed for the field reality of Koraput: offline‑first mobile capture, verifiable evidence, multilingual UX, and policy‑compliant governance. For citizens, this means issues are seen and solved faster. For the administration, this means one pane of glass for planning, interventions, and reporting. For shareholders, Koraput establishes a replicable GovTech product line with clear unit economics and a pathway to state‑wide scale.

1. Problem Landscape & Objectives

Fragmented data and delayed decisions. Departmental apps and paper registers create blind spots. District officers rely on monthly summaries that average out local spikes (absenteeism, meal gaps, water outages), resulting in slow, blunt interventions.

Weak evidence and audit friction. Photos without metadata, unverifiable timestamps, and non‑standard forms complicate internal audit and external reviews. Grievances go into queues without SLA discipline or cross‑department ownership.

Capacity and change fatigue. Field staff juggle multiple apps and forms, each with its own login and workflow. Training is episodic; adoption falls off after the first quarter.

Program objectives.

  1. Unify priority use‑cases (education, nutrition, grievances, infrastructure) on one backbone.
  2. Verify with evidence at the point of capture (GPS, time, photo hash, role identity).
  3. Predict risks early (dropouts, meal‑quality anomalies, facility breakdowns).
  4. Enforce SLAs with clear RACI and automated escalations.
  5. Report to district/state/central formats with zero manual collation.

2. Functional Coverage

2.1 Education

  • Attendance for students and teachers with device‑time, geo‑fence, and liveness prompts.
  • Performance inputs (weekly tests, foundational literacy/numeracy checkpoints).
  • Interventions (home visits, parent meetings) logged with outcomes.

2.2 Nutrition (Mid‑Day Meal)

  • Menu & quantity capture with photo evidence; AI checks for portion/consistency.
  • Supply & stock tracking (rice, pulses, oil, eggs, milk); variance alerts vs. planned norms.
  • Quality flags with escalation to block‑level food inspectors.

2.3 Infrastructure

  • Facility uptime: water, electricity, toilets, classroom condition, kitchen sheds.
  • Ticketing to Works/Panchayat departments with evidence and SLAs.
  • AMC registry for recurring services and vendor compliance.

2.4 Grievance Redressal

  • Unified intake (app, helpline, kiosk, WhatsApp).
  • Categorization by department; SLA bands (T+1 triage, T+7 resolution).
  • Citizen feedback loop with satisfaction score and reopened‑case logic.

3. Technical Architecture

3.1 Capture & Identity

  • Offline‑first Mobile App (Android, multilingual): queues form submissions; conflict‑aware sync.
  • Strong identity: device binding, role‑based login, optional Aadhaar‑based e‑KYC for staff where policy allows.
  • Evidence seal: GPS, device time, EXIF, and SHA‑256 image hash attached to each record; tamper‑evident.

3.2 Data Platform

  • Event pipeline: Mobile → Edge buffer → HTTPS/MQ ingestion with back‑pressure controls.
  • Operational store (PostgreSQL) for transactions; Object store for images/docs.
  • Analytics lake/warehouse for trend analysis and ML features; partitioned by block/GP/school.

3.3 ML Services

  • Attendance anomaly detection (spikes, improbable patterns, cross‑correlation with holidays/exams/weather).
  • MDM quality sentinel: image heuristics for portion and temporal consistency; vendor/supplier risk scoring.
  • Dropout risk score combining attendance, performance micro‑assessments, and grievance context.
  • Facility failure prediction: simple survival/propensity models on past outages and AMC cycles.

3.4 Security, Privacy, and Compliance

  • RBAC with least privilege; audit trail on every read/write.
  • Encryption at rest and in transit; key rotation policies.
  • Data minimization and consent flags; retention/archival policies aligned to government norms.
  • Network posture: IP whitelisting for admin console; WAF for public endpoints; automated VAPT cadence.

3.5 Integration

  • State ID registries (student/teacher master), MDM supply systems, grievance helplines, and messaging gateways (SMS/WhatsApp).
  • Export adapters for state dashboards and central schemes to avoid double entry.

4. Governance & Operating Model

4.1 RACI & Committees

  • Steering Committee (Collector/CEO ZP, District Programme Heads, IT) — monthly policy/priority review.
  • Operations Board (BEOs/BDOs/ADMO/WCD leads, Xapny PMO) — weekly performance and bottlenecks.
  • Security & Data Cell (District IT + Xapny security lead) — access reviews, incidents, and audits.

4.2 SOPs & SLAS

  • Education: Daily attendance close by 11:30 AM; auto‑escalation if <85% data coverage in a cluster.
  • MDM: Menu entry before serving; post‑serving photo within 20 minutes; variance alerts >10% trigger BEO check.
  • Grievances: T+1 triage, T+3 interim response, T+7 resolution; automatic escalation to next authority on breach.
  • Infrastructure: Ticket categorization (critical/major/minor); response MTTR targets (24/72/120 hrs).

4.3 Adoption & Training

  • Role‑based curricula: teachers/cooks/cluster coordinators/block officials.
  • In‑app guides and micro‑videos; quarterly refreshers.
  • Champion network per block; usage dashboards visible to leadership.

5. KPIs & Reporting

Coverage & Timeliness

  • Data coverage by cluster/school (daily %), on‑time submission rate, app login cadence.

Service Quality

  • Attendance stability, MDM variance, grievance SLA compliance, facility uptime.

Impact & Equity

  • Dropout risk distribution, resolution equity (urban vs. remote), repeat‑grievance rate.

Reporting Cadence

  • Daily ops digest to BEO/BDO; weekly block leaderboard; monthly district governance pack.
  • Exportable MoPR/State templates; evidence bundles for audits.

6. Implementation Plan

Phase 0 – Blueprint & Readiness (3–4 weeks)

 Stakeholder mapping, data audit, baseline survey tools, infra sizing, RACI finalization, and communication plan.

Phase 1 – Education + MDM Pilot (8–10 weeks)

 10–15 schools/hostels across varied geos; offline sync tests; KPI baselines vs. pilot deltas; first steering review.

Phase 2 – Grievances + Infrastructure (6–8 weeks)

 Unified intake, SLA engine, ticket routing to Works/Panchayat; Works vendor onboarding; AMC registries.

Phase 3 – District‑Wide Scale (12–16 weeks)

 Wave‑based rollout; hypercare; training completion >90%; dashboard public releases where policy allows.

Phase 4 – Predictive & Policy (ongoing)

 Dropout sentinel rollout; MDM supplier risk; infrastructure failure predictions; quarterly policy recalibration.

7. Risk Register & Mitigations

  • Connectivity Gaps: Offline‑first, store‑and‑forward; SMS fallback; periodic bulk sync windows.
  • Device Sharing & Credential Risk: Device binding, session expiry, and anomaly lockouts; periodic credential refresh.
  • Adoption Drop‑off: UI simplification; incentive‑aligned KPIs; champion follow‑ups; voice prompts in local languages.
  • Data Quality Gaming: Liveness checks, randomization of prompts, cross‑checks between attendance and MDM servings.
  • Privacy Concerns: Strict purpose limitation; opt‑out for non‑mandatory fields; transparent policy screens.
  • Inter‑department Coordination: Shared SLA board; Collector‑level escalation for persistent breaches.

8. Commercials & Value Realization

  • Model: District subscription with per‑institution pricing; optional O&M for field devices.
  • ROI Drivers: Reduced leakages in MDM, faster ticket resolution, early dropout interventions, fewer audits/inspections required due to verifiable evidence.
  • Payback: 9–15 months depending on coverage and leakage baselines; additional savings via shared infra with state IT.

9. Strategic Fit for Xapny

Koraput demonstrates Xapny’s ability to productize governance: one backbone serving multiple public‑service verticals, each with domain‑specific workflows and shared AI services. This positions Xapny as a GovTech scale partner for Odisha’s districts and other states, complementing our PSU portfolio (IOCL, NALCO, MCL) with direct citizen‑outcome programs.

10. Next Steps

  1. Approve Phase‑1 pilot cohort and SLAs.
  2. Provision district cloud/on‑prem env; whitelist messaging gateways.
  3. Conduct trainer‑of‑trainers; launch communications and helplines.
  4. Start Phase‑1 rollout with weekly Ops Board reviews and public KPI dashboards (where policy permits).

Conclusion. Koraput’s unified monitoring platform converts fragmented reporting into verifiable, actionable, and predictive governance. It is engineered for India’s field reality and built for scale—district by district—creating durable value for citizens and a defensible, replicable business for Xapny.