Humanitarian Programme Management Platform

Outcomes

Context

Multi-country programme operations with partner organizations, auditors, and field teams that cannot assume reliable connectivity.

Challenge

Programme teams ran planning, partner reporting, travel, audits, and last-mile logistics across fragmented tools. Field staff needed offline collection; headquarters needed a coherent picture of partnerships, results, and risk. Legacy frontends and multiple backends made delivery slow and inconsistent.

Approach

We consolidated delivery into a monorepo with shared UI packages and Docker-based local/prod parity. Backends stayed domain-oriented (programme management, partner reporting, cluster reporting) while frontends reused a common component layer. Offline collection and last-mile tracking were treated as first-class products with explicit sync and API contracts.

Architecture

Django and PostGIS backends, Polymer/Lit frontends in a pnpm monorepo, Nginx edge routing, Docker Compose profiles for selective app stacks, and dedicated offline-collect and last-mile services (including Next.js + Nest for field logistics).

Scope

• —Programme management portals
• —Partner and cluster reporting
• —Offline data collection
• —Last-mile delivery tracking
• —Shared design-system packages
• —Dockerized local and production-like stacks

Constraints

• —Long-lived legacy UI frameworks that still ship production traffic
• —Strict data segregation between programme domains
• —Field connectivity constraints
• —Multiple stakeholder roles with different access needs

Solution highlights

• —Monorepo workspaces for shared frontends and coordinated releases
• —PostGIS-backed geospatial reporting for partner and cluster views
• —Offline-first collect client with sync-oriented API
• —Last-mile application stack separated but integrable with programme APIs

Technologies

Results

• —Single local orchestration path for core programme apps instead of many ad-hoc repos
• —Shared UI packages reduced duplicate components across portals
• —Offline collection and last-mile flows integrated with the same programme data model
• —Clearer ownership boundaries between programme, partner, and cluster reporting stacks

Lessons learned

• —Humanitarian platforms succeed when offline and online are designed together, not bolted on.
• —Shared UI packages matter more than a single mega-frontend when domains must stay separate.
• —Compose profiles are a practical way to keep large suites runnable for new engineers.