KiraFlo

Inspiration

The inspiration for KiraFlo.ai stems from the "Receipt-to-Insight Gap" prevalent among micro-enterprises and individuals in Malaysia. Business owners are often overwhelmed by manual bookkeeping, risking tax non-compliance (LHDN) and missing growth opportunities. Individuals, meanwhile, struggle with debt traps and stagnant savings. We wanted to build an autonomous bridge—a tool that transforms raw, messy physical receipts into actionable financial intelligence, empowering users to take control of their financial destiny.

What it does

KiraFlo.ai is a dual-persona AI agent platform:

• For SMEs: An autonomous bookkeeping engine that categorizes expenses, ensures tax-readiness, and prepares financial reports for loan acquisition.
• For Individuals: A personalized AI financial coach that tracks budgets, analyzes spending habits, and provides tailored saving strategies.

By automating data extraction, we reduce the time spent on manual accounting (we call this \( T_{manual} \)) by a factor of \( \epsilon \) (where \( \epsilon \ll 1 \)):

$$T_{ai} \approx \epsilon \cdot T_{manual}$$

How we built it

KiraFlo.ai is built on a modular architecture leveraging:

• Core Engine: Python with Google Generative AI (Gemini) for advanced receipt data extraction and natural language categorization.
• Frontend: A streamlined interface designed for low-latency user interaction.
• Backend: Scalable API-first architecture, allowing us to manage different "Personas" (Business vs. Individual) within a single session.
• Deployment: Hosted with a focus on token-efficiency to ensure accessibility even under standard quota limits.

Challenges we ran into

• Infrastructure Instability: During the development phase, we encountered significant downtime in the hackathon platform's backend services. We overcame this by focusing on local development and direct integration with Google AI Studio.
• Defining Scope: Balancing a B2B (Business) and B2C (Individual) value proposition required careful architecture to avoid feature bloat, which we resolved by implementing a modular "Persona" system.

Accomplishments that we're proud of

• Market-Ready Logic: We successfully designed a monetization strategy that balances rapid market validation (via a "Founder’s Tier") with long-term subscription scalability.
• Mathematical Optimization: We developed a clear framework to quantify the efficiency gain ($$T_{ai} \approx \epsilon \cdot T_{manual}$$), making our impact measurable and objective for judges.
• Resilience: We maintained project momentum despite infrastructure hurdles, demonstrating our capability to deliver under pressure.

What we learned

• Market Validation First: We learned that a good product is not just about code, but about proving "willingness-to-pay." Our transition from high price points to a "Founder’s Tier" taught us the value of agile pricing strategies.
• AI-First Engineering: We learned to design systems that are token-efficient, ensuring that our AI agents are both cost-effective and highly responsive.

What's next for KiraFlo

• Scaling Validation: We are moving into the "Market Validation Phase," onboarding our first batch of local SMEs to gather real-world usage data.
• Financial Ecosystem: We plan to integrate with local financial institutions to streamline the loan and insurance referral process.
• Predictive Analytics: Expanding our AI agent's capabilities to include predictive cash-flow forecasting, helping businesses prepare for financial gaps before they happen.

Built With

• fastapi
• git
• google-cloud
• googlegemini
• json
• postgresql
• python
• sqlite
• streamlit
• virtual-environment
• vscode