Next Founders Hackathon 2026

Submission: FounderOS — from an idea to something real

Try it!


Live Website Source on GitHub Read the full write-up on Devpost

Submission


Submitted to Devpost as FounderOS. The stack it was built on:

Swift Kotlin Python Node.js DeepSeek API Clerk Vercel

Built with Isabella Wu and Annie Liao; the full credits are on Devpost. Started September 2026.


About the Project


Inspiration: Plenty of students and first-time founders have ideas that could solve real problems, but never take the first step — no mentor, no funding, no idea where to start. We didn't want another chatbot that answers questions. We wanted something that turns an overwhelming journey into a series of small, concrete steps.

What it does: You submit an idea and FounderOS returns a structured startup analysis: a name and tagline, the problem, target users, proposed solution, MVP features, business model, validation questions to answer before investing real time or money, one concrete next step, and future features. Signed-in users can save and manage ideas. The point isn't to promise an idea will work — it's to surface the assumptions so you can test them.

How it was built: One shared backend, thin clients. The AI analyzer sits behind a backend API that calls DeepSeek and returns strict JSON, so the iOS app (Swift), the Android app (Kotlin) and the website can all decode the same fields without duplicating business logic. The website uses Clerk for accounts and saved ideas and is deployed on Vercel.

Challenges: Keeping the DeepSeek API key out of the mobile apps. A key embedded in an app can be extracted, so every client talks to our backend, and only the backend holds the key — which also gives us the one place to add authentication, rate limiting and cost controls. The other fight was scope: entrepreneurship covers market research, customer discovery, finance and product. We chose one useful starting point — idea to structured plan to next step — instead of a shallow version of everything.

What I learned: To think in systems. An AI product isn't a model plus a UI; it's a client, a backend, an external provider and a data contract between them, each with its own failure points. Writing the API contract down first is what let three clients build against it at once.

Dev Notes

Architecture

One backend, two thin native clients plus a website. The API contract is the interface between them.

Security Choice

The provider key lives only on the server; clients never see it. Auth, rate limits and usage caps attach there.

Structured Output

The model returns fixed JSON fields, not prose, so every client can decode and display the same analysis.

What's Next

Grow from a plan generator into a guided path: discover, validate, plan, build, launch and improve.


← Back to Hackathons Back to top