Connecting Salesforce, SITS and Canvas for a global digital campus
The University of London's Global Digital Campus programme needed student, enrolment and learning data to move reliably between Salesforce, Tribal SITS and Canvas — in near real time, and without destabilising systems the institution depends on.
- Client
- University of London
- Programme
- Global Digital Campus
- Period
- 2025 – present
- Role
- Integration developer
Delivered by Brad McAllister, founder of Real Code Ltd, while embedded in the institution's engineering team.
01
The challenge
A global digital campus only works if its core systems agree with each other. Student records lived in SITS, customer and pipeline data in Salesforce, and learning activity in Canvas — three platforms with different data models, different update cadences and no shared contract.
The programme needed a synchronisation layer that could move high volumes of changes between them with low latency, recover cleanly from failure, and give the engineering team visibility when something went wrong.
02
What was delivered
A decoupled integration ecosystem built on Azure Service Bus, with a Salesforce gRPC architecture integrated with Azure Common Data Model for robust, near-real-time synchronisation across the institution.
Resilient serverless microservices in Python on Azure Functions act as the intermediary layer — orchestration engine and data ingest between Service Bus and Cosmos DB, scaling with load rather than with headcount.
A custom administrative dashboard in React and Next.js gives the engineering team real-time observability over middleware containers and direct management of Service Bus queues and dead-letter queues — failures are recovered through tooling, not guesswork.
03
The controls
Message persistence and dead-letter handling mean failures surface as managed exceptions rather than silent data loss — the same pattern NetEvolution applies to agent action logging.
A centralised error-logging and telemetry framework routes exceptions to ServiceNow, Cosmos DB and Azure Log Analytics. It was adopted by multiple engineering teams beyond its original scope and measurably reduced mean time to resolution.
04
Why it matters for AI
This is the layer agentic systems stand on. An AI agent reading student records or updating a CRM is only as trustworthy as the integration underneath it — durable queues, schema discipline and an audit trail are prerequisites, not extras.
The engagement is current and ongoing, which is why we describe architecture and controls precisely and quote no invented efficiency figures.