Sofia

Inspiration

Every year in April, Rwanda experiences devastating floods that wash away roads, isolating millions from essential medical care. Building long-term infrastructure would cost billions and take decades—until a company called Zipline introduced drone delivery, solving the problem in under a year. That’s leapfrog technology: skipping outdated infrastructure by jumping straight to scalable innovation.

This sparked a big question for us:

What if we could turn the Zipline story into a replicable formula for all sectors, not just healthcare?

We envisioned a platform where any entrepreneur in the world—regardless of their access to capital or formal education—could team up with an AI that understands their local context, knows every successful business strategy ever implemented, and provides step-by-step guidance to scale their business.

What it does

We built Sofia—an AI-powered Business Empowerment System (BES) for founders in low-resource environments. Sofia is your:

🤝 Business coach

🧭 Strategy advisor

📊 Market analyst

💬 Conversational partner

Users input their product, team, market, and goals. Sofia then generates a personalized, dynamic roadmap with milestones, marketing plans, financial forecasts, legal guidance, and more—tailored to their local reality.

Sofia also:

• Provides competitor insights from scraped market data
• Suggests funding opportunities and mentors
• Offers a mobile-friendly, multilingual interface (with voice input for literacy-limited users)
• Enables users to iterate with a human-AI collaborative dashboard

How we built it

We divided development into four hackathon phases over 36 hours: 1 - System Setup: Next.js frontend, FastAPI backend, PostgreSQL database 2 - AI Agent Integration: GPT-4-turbo / Mistral 7B for business planning, LangChain for prompt orchestration 3 - Dashboard & UX: Interactive business plan interface, real-time tracking, mobile-optimized UI 4 - Testing & Deployment: Vercel (frontend), Fly.io (backend), Pinecone for vector search, multilingual + voice UX

Challenges we ran into

• Designing for low-connectivity and low-literacy environments while maintaining a powerful AI experience
• Tailoring the AI agent to balance global strategy knowledge with localized business context
• Building a tool that’s intuitive and empowering for non-technical users without overwhelming them
• Ensuring scalability of personalized outputs while maintaining AI accuracy

What we learned

• Great AI is only as impactful as its human interface—UX for underserved users is a unique design challenge
• The most transformative ideas often come from asking: “What if we built this for the people most often left behind?”
• Real innovation doesn’t just automate—it empowers!

Built With

• cohere
• cohere-(exploration)-vector-db:-faiss-/-pinecone-database:-postgresql-platforms:-vercel
• css
• faiss
• fastapi
• figma
• fly.io-libraries-&-tools:-langchain
• gemini
• huggingface-transformers
• javascript
• mistral-7b-(via-huggingface)
• next.js
• pinecone
• postgresql
• postman
• python
• railway
• sql
• sql-frontend:-next.js
• tailwind
• tailwind-css-backend:-fastapi-ai-models:-openai-gpt-4-turbo
• typescript
• vercel
• whisper-api