AegisFi — Autonomous AI Micro-Underwriting Protocol

• 

  AegisFi Autonomous Underwriting Console with Live Gemini AI Risk Memo and Tri-Agent consensus telemetry.

• 

  Tri-Agent Pipeline: Real-time autonomous decisions from Sentinel, Quant, and Compliance Auditor agents.

• 

  Dynamic Risk Mitigation: Real-time remediation coaching and credit repair roadmap powered by Gemini.

• 

  Offline-first retro USSD interface (*384#) with Swahili and English localization for feature phones.

• 

  Audit-ready Institutional Credit Memorandum generated instantly for tier-1 partner banks.

• 

  Standardized credit memo featuring factor decomposition, DSCR analysis, and cryptographic signatures.

Inspiration

What it does

💡 Inspiration

Across Sub-Saharan Africa and emerging markets, over 44 million informal micro-merchants (Dukas, Boda operators, kiosk vendors) drive 80% of employment but face a crippling $330B credit gap. Traditional financial institutions reject them because they lack formal credit histories and bank statements. We built AegisFi to turn raw, informal mobile money cashflows (M-Pesa) into high-fidelity, institutional-grade creditworthiness in real time.

⚙️ What It Does

AegisFi is an autonomous underwriting protocol and credit intelligence engine:

1. Tri-Agent Consensus Pipeline: Autonomous agents (Fraud Sentinel, Quant Risk Modeler, Regulatory Auditor) simultaneously evaluate transaction streams in under 50ms.
2. Gemini AI Intelligence: Generates institutional-grade executive risk memorandums and actionable financial remediation coaching tailored for informal merchants.
3. Emerging Market Rails: Features an offline-first Retro USSD simulator (*384#) and dual-language localization (English & Swahili) for feature-phone access.
4. Daraja B2C Webhook & Telemetry: Simulates real-time automated fund disbursement and generates audit-ready PDF Credit Memorandums for tier-1 partner banks.

🛠️ How We Built It

• Backend & Core Engine: Python 3.11+, FastAPI, Uvicorn, Pydantic.
• AI Reasoning: Google Gemini API (Gemini 1.5/2.5 Flash) via native REST integration for ultra-low-latency structured reasoning.
• Mathematical Underwriting: Sigmoid-normalized scoring matrix (300–850 scale), Debt Service Coverage Ratio (DSCR), and cashflow volatility weighting.
• Interface: High-density telemetry dashboard with real-time consensus visualization, USSD simulator, and cryptographic verification logs.

💪 Accomplishments that we're proud of

• Achieving sub-50ms deterministic scoring while integrating complex, multi-agent AI reasoning.
• Designing authentic emerging-market features, including offline USSD simulation and genuine Swahili financial remediation.
• Delivering a production-ready, institutional-grade repository with complete architectural documentation.

🔮 What's next for AegisFi

• Deploying live integrations with Safaricom's production Daraja API sandbox.
• Expanding on-device local SLM inference for zero-connectivity edge underwriting in remote rural areas.
• Partnering with microfinance institutions (MFIs) across Kenya, Tanzania, and Rwanda. ## How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for AegisFi — Autonomous AI Micro-Underwriting Protocol

Built With

• fastapi
• google-gemini
• pydantic
• python
• rest-api
• uvicorn