Narration script
iECHO technical overview: Engineered for population scale
About 9 minutes at a natural speaking pace
-
-
Opening
This is a technical overview of iECHO, the platform behind the global ECHO community. It serves more than 700,000 active users and more than 1,200 hubs in more than 70 countries, in more than 25 languages. It's open and API-first, built on cloud-native microservices, secured with a Zero Trust-inspired approach and ready for AI. Let's look at how it's engineered for population scale.
-
-
-
Opening
Eight principles shape every part of iECHO. It's API-first, open and interoperable, and cloud-native. Every environment is provisioned from code. It's designed mobile-first, for low bandwidth and shared devices. It's secure by default and private by design, with no patient data. And it's observable, with metrics, logs and traces on every request.
-
-
-
Opening
Here's the reference architecture, from top to bottom. People reach iECHO through the web app, built in React and TypeScript, mobile apps for Android and iPhone, WhatsApp, partner apps and admin consoles. Every request passes through the edge, with a global CDN, a web application firewall and DDoS protection, then an API gateway that validates tokens, limits rates, routes and versions requests, and traces every call.
-
Step 1
Behind the gateway, domain microservices in Java, Spring Boot and Python each own one part of the ECHO Model, from identity and hubs to sessions, attendance, certificates and AI. They talk to each other through an asynchronous messaging backbone of events, queues and background jobs.
-
Step 2
They share a common platform for master data, taxonomy, identity, localization, search and analytics. Data lives in MySQL for the relational core, DynamoDB for high-volume events, Redis for caching, object storage for media and a warehouse for analytics. It all runs on AWS, in containers on Kubernetes, provisioned by Terraform across multiple availability zones.
-
Step 3
Alongside every layer, we observe and secure the platform with CloudWatch, Elastic APM, Grafana and Loki, GuardDuty and Inspector, with a public status page. And iECHO integrates with the tools our community already uses, from Zoom and WhatsApp to Salesforce, Freshworks, Wistia and Mapbox.
-
-
-
Opening
iECHO is built as domain-driven microservices. Each service owns one part of the ECHO Model, from identity and roles to programs, sessions, attendance and certificates, content and discussions, notifications, telemetry and AI, and each owns its own data and scales on its own.
-
Step 1
Every service follows the same anatomy: a versioned API described in OpenAPI, stateless logic, its own data store, events it publishes and subscribes to, a container with health checks and an autoscaling policy, and telemetry out of the box.
-
Step 2
Services work together through events. When a session ends, one event fans out: attendance finalizes the register, certificates are issued, participants are thanked on email and WhatsApp, dashboards update and AI drafts the session recap. One event, five reactions, and no service waits on another.
-
-
-
Opening
iECHO is API-first. Our own apps use the same APIs we open to partners: REST and JSON over HTTPS, described in OpenAPI and versioned. A partner app with the right OAuth scope can ask for a program's upcoming sessions, in the participant's language.
-
Step 1
It gets back clean, structured data, paginated with a cursor and traced end to end, with the same taxonomy of focus areas used across the platform.
-
Step 2
And when something happens, like a session ending, iECHO sends a signed webhook in real time, so partners never have to poll.
-
-
-
Opening
Every request is authenticated and authorized. A user signs in with a password and a one-time code, and the identity service issues a signed, short-lived token. The API gateway verifies its signature, expiry and scopes on every call, and the domain service checks the user's role in that specific hub and program before returning only what the role allows.
-
Step 1
Access control is multi-tenant. One identity spans every hub a person works in, from a ministry to state hubs, programs and participant groups, with a role in each. Hub admins configure the hub, hub members and experts work only within the programs they're responsible for, and participants join and discuss. It's a Zero Trust-inspired model: every request verified, least privilege everywhere and internal tools behind a VPN.
-
-
-
Opening
Data and telemetry are at the core. Events flow in from every domain service, from session telemetry in Zoom, from surveys and forms, and from the web, mobile and WhatsApp clients. They're ingested in real time through event streams and change data capture, with schema checks that stop bad data at the door.
-
Step 1
The warehouse keeps data in three tiers: raw and immutable, then curated, then de-identified. Between them sits a privacy gate that de-identifies and aggregates, so the data we share and learn from is safe by construction.
-
Step 2
From there, it powers dashboards in Apache Superset and Metabase, maps in Mapbox, open APIs and exports for partners, and features for AI and the Listening Network. And it all shares one master data and taxonomy spine: professions, focus areas, professional IDs, geographies, institutions and languages, so data means the same thing everywhere.
-
-
8. Built to scale
30 s
-
Opening
iECHO is built to scale from one program to a whole country. Services are stateless and scale out pod by pod. Caching at every layer keeps reads cheap, queues absorb bursts, write-heavy data like telemetry and attendance is partitioned in DynamoDB, and there's no single point of failure across availability zones. Picture a ministry launching a national program: traffic jumps fifty times in ten minutes, the platform scales out to meet it, and latency holds steady.
-
-
9. Security in depth
58 s
-
Opening
Security is built in depth, in five layers, and mapped to the NIST Cybersecurity Framework 2.0. iECHO is ISO 27001 certified and compliant with GDPR, CCPA, DPDPA and FISMA. At the edge, a web application firewall, DDoS protection and TLS for every connection.
-
Step 1
In the network, private subnets and segmented networks, with internal tools reachable only over a VPN.
-
Step 2
For identity, role-based, least-privilege access, two-factor authentication for admins and short-lived signed tokens, with every request verified.
-
Step 3
In the application, a secure development lifecycle, dependency and container scanning, automated tests on every change and secrets kept out of code.
-
Step 4
And at the core, the data: encrypted at rest and in transit, minimized, and with no patient data, by design.
-
Step 5
Across every layer, we detect and respond around the clock, with GuardDuty, continuous scans from Inspector, centralized logs, a formal incident response plan and live status at status.iecho.org.
-
-
-
Opening
We ship small changes often, and every one is tested, scanned and reversible. Code is peer reviewed in GitHub. Jenkins builds it and runs unit tests, then end-to-end tests in Selenium and security scans. Each build becomes an immutable Docker image, Terraform provisions the infrastructure as code, and Kubernetes rolls it out with zero downtime, with health checks that roll back automatically.
-
Step 1
Every release is promoted from development to staging to our beta partners and then to production, through environments built from the same code. Nothing ships without passing its tests and scans, and every release comes with published release notes.
-
-
-
Opening
The stack is modern and open, built from proven, widely adopted technologies: React and TypeScript on the front end, with apps for Android and iPhone; Java, Spring Boot and Python services; MySQL, DynamoDB and Redis for data; and AWS, Kubernetes, Docker, Terraform, Jenkins and GitHub for cloud and delivery. We observe it with CloudWatch, Elastic APM, Grafana and Loki, analyze with Apache Superset, Metabase and Mapbox, and integrate with Zoom, WhatsApp, Salesforce and leading AI models.
-
-
-
Opening
Because iECHO is API-first and data-rich, it's AI-ready by design. The core platform's applications, microservices and data, from didactics and recordings to cases and discussions, feed a family of AI platform services through open APIs and de-identified data: collective intelligence, generative AI, search and recommendations, language understanding, conversational assistants, classification, translation, quality monitoring, trust and integrity, PHI and PII detection, and trend analysis.
-
Step 1
They run on a shared AI infrastructure: a model gateway across Gemini, Claude and open-weight models, a knowledge graph and vector index, speech and vision, and evaluations with human feedback. Our iECHO + AI presentation goes deeper.
-
-
-
Opening
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. We're looking for partners in cloud and AI infrastructure, in engineering collaboration on open-source components and standards, and in research, from evaluations in low-resource languages to the ECHO Listening Network. Let's build population-scale learning infrastructure together.
-