iECHO: Engineered for population scale

The open, API-first platform behind the global ECHO community: cloud-native microservices, Zero Trust-inspired security and AI built in.

Active users
700,000+
Hubs
1,200+
Countries
70+
Languages
25+
Data and cloudMySQL, DynamoDB, Redis, AWS MicroservicesDomain services on Kubernetes Open APIsGateway, auth, rate limits, webhooks ExperiencesWeb, mobile, WhatsApp, partner apps

Architecture principles

Eight decisions that let one platform serve a single program or an entire country.

  1. 01

    API-first

    Every capability is a versioned, documented API, for our own apps and for partners

  2. 02

    Open and interoperable

    Open-source components and open data standards, with no lock-in

  3. 03

    Cloud-native

    Stateless, containerized services that scale out on their own

  4. 04

    Infrastructure as code

    Every environment provisioned, versioned and reproduced from code

  5. 05

    Mobile-first

    Designed for phones, low bandwidth and shared devices

  6. 06

    Secure by default

    Zero Trust-inspired access and encryption everywhere

  7. 07

    Privacy by design

    Data minimization, no patient data and compliance with GDPR, CCPA and DPDPA

  8. 08

    Observable

    Metrics, logs and traces on every request, from day one

Reference architecture

Observe and secure

  • CloudWatchMetrics and alarms
  • Elastic APMApplication traces
  • Grafana + LokiCentralized logs
  • GuardDuty24/7 threat detection
  • InspectorContinuous vulnerability scans
  • status.iecho.orgPublic status page

Clients

iECHO webReact · TypeScriptMobile appsAndroid · iOSWhatsAppMessaging channelPartner appsOpen APIsConsolesAdmin and analytics

Edge

Global CDNWeb application firewallDDoS protectionTLS everywhere

API gateway

Token validationRate limitingRouting and versioningRequest tracingREST · OpenAPIWebhooks

Domain microservicesJava · Spring Boot · Python

IdentityHubsProgramsSessionsSchedulingAttendanceCertificatesContentDiscussionsSurveys and formsNotificationsSearchTelemetryAI services

Asynchronous messaging backbone: events, queues and background jobs

Shared platform

Master dataTaxonomyIdentityLocalizationSearchAnalytics

Data

MySQLRelational coreDynamoDBHigh-volume eventsRedisCache and sessionsObject storageMedia and recordingsWarehouseAnalytics

Cloud

AWSKubernetesDockerTerraformMultiple availability zonesAutoscaling

Integrations

  • ZoomSessions and attendance
  • GupshupWhatsApp messaging
  • SalesforcePartner relationships
  • FreshworksSupport desk
  • WistiaVideo hosting
  • MapboxMaps and geography
  • CalendarsInvitations everywhere

Domain-driven microservices

Each service owns one part of the ECHO Model and its own data, and scales on its own.

  • Identity and access
  • Program delivery
  • Learning and community
  • Engagement
  • Insight and AI

Anatomy of a service

  1. ContractA versioned REST API, described in OpenAPI
  2. LogicSpring Boot or Python, stateless
  3. DataIts own store, never shared directly
  4. EventsPublishes what happened; subscribes to what it needs
  5. RuntimeA container with health checks and an autoscaling policy
  6. TelemetryMetrics, logs and traces out of the box

event session.ended
session ses_9f2c · attended 368

  • Attendancefinalizes the register
  • Certificatesissues verifiable certificates
  • Notificationsthanks participants on email and WhatsApp
  • Telemetryupdates the dashboards
  • AI servicesdrafts the session recap

One event, five reactions, no service waiting on another.

Identity Profiles Rolesand permissions Hubs Programs Scheduling Sessions Attendance Certificates Content Discussions Surveysand forms Search Notifications Telemetry AI services

API-first and open by design

Our own apps use the same APIs we open to partners, so every capability is integration-ready from day one.

  • REST and JSON over HTTPSDescribed in OpenAPI, versioned in the path
  • OAuth 2.0 scopesPartner apps get only the access they're granted
  • Built for scaleCursor pagination, filtering and rate limits
  • Real-time webhooksSigned events, so partners never have to poll
  • Open data standardsExports and a shared taxonomy that travel between systems
GET /v1/programs/4821/sessions?from=2026-11-01
Authorization: Bearer eyJhbGciOiJSUzI1NiJ9…
Accept-Language: mr
200 OK  · 38 ms · trace 7f3a9c
{
  "data": [{
    "id": "ses_9f2c",
    "title": "Dengue case management",
    "starts_at": "2026-11-04T09:30:00+05:30",
    "focus_areas": ["infectious-disease"],
    "attendance": { "registered": 412, "attended": 368 }
  }],
  "next_cursor": "c3VtbWl0"
}
POST https://partner.example.org/hooks/iecho
X-iECHO-Signature: sha256=7d3e4b…
{ "event": "session.ended", "session": "ses_9f2c" }

Illustrative request. Endpoints and figures are examples.

Identity and access

App Identity service API gateway Domain service 1 Sign in: password + one-time code 2 Signed, short-lived token + refresh token 3 GET /v1/programs/4821 + bearer token 4 Verify signature, expiry and scopes 5 Request + verified identity 6 Check role in this hub and program 7 200 OK: only what the role allows

OAuth 2.0 and OpenID ConnectTwo-factor authenticationPasswords hashed, never visibleRevocable sessions

One identity, a role in every hub

  1. Ministry or national institute
  2. State hub
  3. Program
  4. Participant group
Configure hubCreate programsRun sessionsShare contentView reportsJoin and discuss
Hub Admin
Hub Member
Subject matter expert
Participant

Across the hubScoped to their programs

Zero Trust-inspired: every request is verified, least privilege everywhere, and internal tools sit behind a VPN.

Data and telemetry at the core

Every session, click and survey becomes trustworthy, privacy-safe data, ready for dashboards, partners and AI.

Sources

Domain servicesOperational events

ZoomSession telemetry

Surveys and formsFeedback and case forms

ClientsWeb, mobile and WhatsApp

Ingest

Event streamsIn real time

Change data captureFrom every data store

Schema checksBad data stopped at the door

Warehouse

RawImmutable, as it arrived

CuratedCleaned, joined, typed

Privacy gate: de-identify and aggregate

De-identifiedSafe to share and to learn from

Serve

DashboardsApache Superset · Metabase

MapsMapbox

PartnersOpen APIs and Excel exports

AIFeatures for models and the Listening Network

Master data and taxonomy

ProfessionsFocus areasProfessional IDsGeographiesInstitutionsLanguages

Built to scale, from one program to a whole country

  • Stateless servicesScale out horizontally, pod by pod, on demand
  • Caching at every layerCDN, Redis and read replicas keep reads cheap
  • Queues absorb burstsHeavy work moves to background workers
  • Write-heavy data, partitionedDynamoDB for telemetry and attendance at volume
  • No single point of failureSpread across multiple availability zones
  • New regions from codeTerraform stands up a full environment in hours
A national launch, absorbedRequestsRunning podsp95 latency
0 min15 min30 min45 min60 min10×30×50× 50× traffic in 10 minutes

Illustrative load profile.

Security in depth

Five layers of defense, watched around the clock, mapped to the NIST Cybersecurity Framework 2.0.

Edge

  • Web application firewall
  • DDoS protection
  • TLS for every connection
  • Rate limits at the gateway

Network

  • Private subnets and segmented networks
  • Internal tools reachable only over VPN
  • Security groups that allow the minimum

Identity

  • Role-based, least-privilege access
  • Two-factor authentication for admins
  • Short-lived, signed tokens
  • Every request verified: Zero Trust-inspired

Application

  • Secure development lifecycle and code review
  • Dependency and container scanning
  • Automated tests on every change
  • Secrets kept out of code

Data

  • AES encryption at rest
  • Encryption in transit
  • Data minimization
  • No patient data, by design

Detect and respond

  • 24/7 threat detection with GuardDuty
  • Continuous scans with Inspector
  • Centralized logs with Grafana and Loki
  • A formal incident response plan
  • Live status at status.iecho.org
Edge Network Identity Application Data

  • ISO 27001

    Certified

  • GDPR

    Compliant

  • CCPA

    Compliant

  • DPDPA

    Compliant

  • FISMA

    Compliant

  • 508/ADA

    Compliant

  • WCAG

    Compliant

From commit to production, continuously

Small changes, shipped often, each one tested, scanned and reversible.

  1. Code

    GitHub

    Pull requests and peer review

  2. Build

    Jenkins

    Compile and unit tests

  3. Test

    Selenium

    End-to-end and security scans

  4. Package

    Docker

    Immutable, versioned images

  5. Provision

    Terraform

    Infrastructure as code

  6. Deploy

    Kubernetes

    Rolling, zero-downtime releases

  7. Observe

    CloudWatch · APM

    Health checks and automatic rollback

DevelopmentStagingBeta partnersProduction
  • Every environment identical, built from the same code
  • No release without passing tests and scans
  • Zero-downtime deploys, with one-step rollback
  • Release notes published for every release

A modern, open stack

Proven, widely adopted technologies, chosen so partners can build on them and engineers can join quickly.

Frontend

ReactTypeScriptAndroidiOS

Services

JavaSpring BootPythonREST · OpenAPI

Data

MySQLDynamoDBRedisObject storage

Cloud and DevOps

AWSKubernetesDockerTerraformJenkinsGitHub

Quality and observability

SeleniumCloudWatchElastic APMGrafanaLoki

Security

GuardDutyInspectorWAFVPNTwo-factor auth

Analytics and maps

Apache SupersetMetabaseMapbox

Integrations and AI

ZoomWhatsApp · GupshupSalesforceFreshworksWistiaGeminiClaude

AI-ready by design

Core iECHO platform

Applications

WebMobilePartner apps

Microservices

Domain services for every part of the ECHO Model

Data

DidacticsProgramsRecordingsCasesTelemetryDiscussions

ECHO AI platform services

  • Collective intelligenceDiscovering knowledge across discussions
  • Generative AIContent, topics and graphics
  • AI-powered searchDiscovery in every language
  • Recommendation engineA personal learning journey
  • Language understandingQuality, sentiment and equity of conversation
  • Conversational AIAssistants for support and onboarding
  • Classifier enginePrograms and content tagged to the taxonomy
  • Translation engineEvery experience in every language
  • Quality monitoringTrends in program performance
  • Trust and integrityAnomalies, fraud and misinformation
  • PHI and PII detectionProtected health information, flagged
  • Trend analysisEarly signals from cases across the network

AI infrastructure

Model gatewayRouting, caching, failoverModelsGemini · Claude · open-weightKnowledgeKnowledge graph and vector indexSpeech and visionTranscription and videoQualityEvals and human feedback (RLHF)

Go deeper in the iECHO + AI presentation →

Build population-scale learning infrastructure with us

iECHO already connects more than 700,000 people in more than 70 countries. With the right partners, it can reach every health worker, teacher and expert who needs it.

  1. 01

    Cloud and AI infrastructure

    Compute, models and credits to serve every country, in country

  2. 02

    Engineering collaboration

    Open-source components, interoperability standards and shared tooling

  3. 03

    Research partnerships

    Evaluations in low-resource languages and the ECHO Listening Network

Security at iECHO status.iecho.org

Project ECHO and iECHO are registered trademarks of The University of New Mexico

Use the arrow keys or click to move through the slides

All slides

  1. 1Engineered for population scale
  2. 2Architecture principles
  3. 3Reference architecture
  4. 4Domain-driven microservices
  5. 5API-first and open by design
  6. 6Identity and access
  7. 7Data and telemetry
  8. 8Built to scale
  9. 9Security in depth
  10. 10From commit to production
  11. 11A modern, open stack
  12. 12AI-ready by design
  13. 13Build population-scale infrastructure with us