Raincloud

Architecture
Main Screen
Tasks Screen
Notifs Screen
Profile Screen

Inspiration

We were inspired by one simple gap: AI tools are powerful, but meaningful workflows still feel desktop-only. People can message AI from a phone, but they still can’t reliably get real, finished outputs without opening a laptop. Raincloud was built to make mobile users first-class builders.

What it does

Raincloud gives users desktop-grade execution from their phone.
A user describes a task, Raincloud clarifies missing details, presents a clear execution plan, runs the task in the cloud, and delivers a finished artifact.

Examples:

Upload a chapter -> receive an audiobook sample
Upload messy data -> receive cleaned CSV + validation report
Request a repo change -> receive a GitHub PR link + summary

How we built it

We built Raincloud as a mobile control plane with cloud execution:

Mobile app (Expo): task composer, clarification thread, plan review, approvals, status, artifacts
Backend API: task orchestration, planning, permission checks, runtime/cost estimation, lifecycle management
Queue + workers: approved tasks run asynchronously in ephemeral cloud jobs
Storage + notifications: artifact delivery and push updates when runs complete
Integrations: GitHub and task-specific external APIs for media/data workflows

Challenges we ran into

Designing a mobile UX that stays simple while supporting complex execution
Building reliable async orchestration across planning, queueing, execution, and delivery
Making plan generation specific enough to be actionable for very different task types
Handling failures gracefully so users still get useful outcomes and next steps
Keeping execution bounded with clear limits, estimates, and transparent progress

Accomplishments that we're proud of

Shipped an end-to-end MVP that turns phone prompts into real cloud-executed artifacts
Proved a differentiated workflow: clarify -> plan -> approve -> run -> deliver
Enabled both consumer and developer use cases in one product
Built trustworthy execution with user-visible plans, estimates, and result summaries
Delivered a demo that judges can understand quickly and verify through outputs

What we learned

Users value predictability and clarity as much as raw AI capability
Artifact-first outcomes are far more compelling than chat-only responses
Mobile-first doesn’t mean lightweight; it means better orchestration and UX decisions
A strong task lifecycle model is essential for reliability in async AI products
Cross-domain pipelines can share one architecture when planning is treated as a first-class layer