Foundry

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Foundry

Foundry — Hackathon Project Overview Inspiration

Modern digital work is fragmented.

Solo creators, students, and small business operators often have to move between spreadsheets, note-taking apps, AI chatbots, project-management tools, and research platforms just to move one idea forward. Instead of spending their time solving problems and executing, they spend a significant amount of time organizing information and maintaining context across different tools.

Foundry was created to solve that friction.

Our vision is to build an AI-native operating workspace where AI does more than answer questions. Foundry understands project context, structures unorganized information, maintains project state, and helps users move from an initial idea to concrete execution.

What Foundry Does

Foundry acts as an AI co-pilot and execution layer for solo builders and small teams.

🧠 Context Synthesis & AI Workflows

Users can provide raw, unstructured ideas, notes, or task briefs. Foundry uses the Gemini API to transform that information into structured workflows and clearer execution paths.

Instead of simply responding with another block of text, the system is designed around turning information into actions.

☁️ Persistent Project Context

Foundry integrates with Firebase to persist:

User sessions Project states AI-generated outputs User information

This allows users to maintain their work instead of starting from zero every time they return.

⚡ Decision Support

Foundry uses structured outputs, including JSON-based responses, to organize information into useful components such as:

Next actions Schedules Resource breakdowns Project information

The goal is to reduce the gap between knowing what to do and actually doing it.

🎯 Low-Friction Workspace

The interface is designed around rapid execution. Users should be able to provide context, interact with the AI, and move toward an actionable plan without having to configure a complicated productivity system first.

How We Built It 🤖 AI Core

Foundry was prototyped and developed using Google AI Studio, with the Gemini API powering its AI capabilities, multi-turn conversations, reasoning workflows, and structured outputs.

🔧 Development

Development was supported by a self-funded one-year Google Pro subscription, providing access to the tools and compute needed for continuous experimentation and testing.

🔐 Backend & Security

Foundry uses Firebase Authentication for user sessions and Cloud Firestore for persistent project and user data.

💻 Frontend

The application uses a responsive interface designed to support both web and Android environments.

Challenges We Faced

Building Foundry wasn't just about connecting an AI API to a chat interface.

Balancing Development with School

As a 14-year-old solo developer, I had to build, debug, test, and refine a full-stack AI product while also managing school and exams.

That forced me to become much more deliberate about prioritization and time management.

Maintaining AI Context

One of the biggest technical challenges was getting Gemini to consistently maintain the required context and produce reliable, structured outputs across multiple stages of a project.

This required continuous experimentation with system instructions, prompting, and structured response formats.

Deployment Constraints

Getting a working prototype was only one part of the challenge. Platform restrictions, billing limitations, age requirements, local SDK issues, and developer-store requirements created additional obstacles.

Instead of abandoning the project, I adapted the development and distribution approach around those constraints.

What We're Proud Of 🚀 An End-to-End AI Workflow

Foundry successfully connects Google AI Studio, Gemini API capabilities, and Firebase Authentication into a unified application.

💰 Lean AI Architecture

The prompt architecture was designed to keep interactions lightweight, with an average target of under 1,300 tokens per interaction and an estimated marginal cost of under $0.001 per workflow, based on the project's current architecture and assumptions.

🛠️ Built Independently

Foundry was taken from an initial concept to a functioning product by a solo developer.

That included designing the product concept, creating the system instructions, integrating the AI layer, designing the database structure, building the interface, testing workflows, and working through deployment challenges.

What I Learned AI-Native Engineering

I learned that building with AI is not simply about putting a chatbot inside an application.

The real challenge is designing the system around the model through:

System instructions Prompt engineering Context management Temperature and generation controls Structured output schemas Cloud Infrastructure

Building Foundry gave me hands-on experience with Firebase Authentication, Firestore, persistent cloud state, and secure API environments.

Resilience & Pragmatism

Perhaps the biggest lesson was learning how to keep moving when the obvious technical path doesn't work.

Local SDK problems, build issues, platform restrictions, and deployment limitations forced me to experiment, simplify, and find alternative approaches instead of stopping at the first obstacle.

What's Next for Foundry 🚀

Foundry is designed to grow beyond its current AI workspace.

🤖 Autonomous Multi-Agent Workflows

Future versions could introduce specialized AI agents capable of handling asynchronous tasks such as research, analysis, and code review.

🌐 Public Web & Native Mobile Releases

The next stage is completing the distribution-ready web and Android versions and making Foundry accessible to real users.

Freemium SaaS Model

The long-term business model is a freemium SaaS platform with capabilities such as:

Team workspaces Higher-level AI capabilities Custom prompt integrations Cloud synchronization Additional advanced features The One-Sentence Story

Foundry is an AI-native execution workspace that turns messy ideas and project context into structured, personalized actions.

Built With

• artificial-intelligence
• gemini-api
• generative-ai
• google-ai-studio
• google-cloud
• llm