Little.PM

Little.PM is the digital platform I recently designed and implemented for the Little Project Manager experiential-education programme. The project started in Dubai and Abu Dhabi and later continued in Kazakhstan. The programme is aimed at children aged 5–18 and uses hands-on mini-projects to develop problem solving, teamwork, planning, creativity, communication, leadership, reflection, resilience and AI literacy. The platform turns approved learning units into live, team-based classroom simulations. A teacher can generate a structured scenario, review it, launch a session, control stages and timed events, organize teams and roles, collect evidence and reflections, and follow progress through dashboards and skills reporting. Students work in a guided project workspace and can use an AI research companion without leaving the approved learning environment. AI was deliberately designed as a co-pilot rather than an autonomous classroom controller. Scenario generation retrieves approved curriculum context and produces a structured blueprint that is schema- and policy-validated before it becomes a simulation; the Django orchestration service remains the authority for launch, pause, stage movement, event release and completion. The research companion is a walled-garden RAG flow: direct identifiers are redacted, only approved school/global knowledge is retrieved, insufficient context is refused, and age-appropriate answers return citations. Assessment and progress-report drafting is grounded in real evidence — student reflections, teacher comments, observations and current skill levels — and remains editable until the teacher explicitly approves it. The implementation uses Next.js 16, Django REST, PostgreSQL with pgvector, Redis/Celery, private S3 storage and a LiteLLM gateway with Cloudflare Workers AI, with optional provider fallbacks. Role-based access, HTTP-only authentication, audit logging and provider/model/latency traceability support a production-shaped architecture while keeping teacher control and student privacy as explicit design constraints.