Proposal

Personal Finance Management App (Vicobi)

You can read the full proposal here: Vicobi Proposal

1. Executive Summary

The Vicobi (Personal Finance Management App) project aims to provide an intelligent, modern, and highly automated personal financial management platform. Vicobi simplifies financial management through 4 main pillars:

1. Smart Recording: Voice input and invoice scanning, eliminating manual data entry barriers.
2. Goal-based Budgeting: Automated creation and flexible management of money jars.
3. Analysis & Control: Provides visual reports and intelligent alert systems.
4. AI Financial Assistant (Chatbot): Integrates AI Chatbot acting as an advisor, supporting inquiries and enhancing financial knowledge.

From a technical perspective, Vicobi is built on a Microservices architecture using .NET Aspire and FastAPI, deployed on AWS Cloud, ensuring flexibility and data security. The development process follows the Agile/Scrum model (2 weeks/sprint during the main development phase), with the goal of completing MVP within 2 months of execution.

2. Problem Statement

Current Problem

In today’s dynamic market, users face difficulties in controlling finances due to “behavioral inertia” — reluctance to manually record each transaction. Existing applications (like Money Lover, Misa Money Keeper) still rely heavily on manual input, causing “input fatigue” and high abandonment rates.

Solution

Vicobi solves the problem through high automation of the data entry process using AWS Cloud and Microservices:

• Core Technology: Integrates AI for Vietnamese voice processing (Voice-to-Text) and detailed invoice recognition (OCR).
• Optimized Architecture: Uses AWS ECS Fargate running Multi-container Task model (combining .NET Backend and AI Service) to reduce infrastructure costs while ensuring seamless communication.
• Modern Frontend: Uses Next.js hosted on Amazon S3 and distributed globally via Amazon CloudFront.

Benefits and Return on Investment (ROI)

The solution provides clear competitive advantages:

• User Value: Reduces over 70% manual operations. Voice recognition accuracy reaches 90% and invoice extraction reaches 80%.
• Economic Efficiency: Maximizes AWS Free Tier usage (S3, CloudFront, Cognito). Lean operating budget around ~$60/month for infrastructure and ~$15/month for AI compute.
• ROI: Expected to achieve ROI within 6–12 months thanks to time savings and increased efficiency.
• Scalability: Microservices architecture ready for Mobile App integration or Open Banking.

3. Solution Architecture

The system is designed with a distributed Microservices model, using API Gateway as the single entry point.

Tech Stack Details:

AWS Workflow:

1. Access: Users access via Route 53, protected by AWS WAF and accelerated by CloudFront.
2. Authentication: Amazon Cognito manages identity and issues JWT Tokens.
3. API Processing: Requests go through API Gateway, connecting securely via AWS PrivateLink to Application Load Balancer (ALB).
4. Compute: ALB distributes load to containers in ECS Fargate (located in Private Subnet).
5. DevOps: CI/CD process fully automated by GitLab, builds images pushed to Amazon ECR and updates tasks on ECS.

4. Technical Deployment

Implementation Phases

The project spans 4 months (including internship):

1. Month 0 (Pre-internship): Ideation and overall planning.
2. Month 1 (Foundation): Learn AWS, upgrade .NET/Next.js/AI skills. Set up VPC, IAM.
3. Month 2 (Design): Design High-level & Detailed architecture on AWS.
4. Month 3-4 (Realization): Coding, Integration Testing, Deploy to AWS Production, set up Monitoring.
5. After Month 5: Research and develop Mobile App.

Detailed Technical Requirements:

• Frontend: Deploy Next.js 16 on S3 + CloudFront. Use Origin Access Control (OAC) to secure bucket.
• Backend:
  • Use .NET Aspire to manage Cloud-native configuration.
  • Database-per-service: PostgreSQL & MongoDB. Elasticsearch for complex transaction search.
  • Background Jobs: Use Hangfire.
• AI Service Pipelines:
  • Voice: Preprocessing with Pydub, PhoWhisper-small Model (VinAI) for Vietnamese.
  • OCR: Amazon Bedrock (Claude 3.5 Sonnet Multimodal) to accurately extract invoice information.
  • Chatbot (RAG): Knowledge Base stored in Qdrant, generates responses via Amazon Bedrock (Claude 3.5 Sonnet).
• Security:
  • Data encryption in transit (HTTPS/TLS 1.2+) and at rest (AES-256).
  • Secrets management not deeply integrated (currently at MVP level), will upgrade to AWS Secrets Manager in the future.

5. Timeline & Milestones (Sprints)

The main execution phase is divided into 4 Sprints:

• Sprint 1: Core Foundation
  • Authentication (Cognito), Wallet Management, Spending Jars.
• Sprint 2: Core Features
  • Transactions (CRUD), AI Voice Processing.
• Sprint 3: Analytics
  • Reports/Charts, Notification System (SES), Message Broker.
• Sprint 4: Stabilization
  • Integration Testing, UI Refinement, Deploy to AWS ECS & CloudFront.
• Testing & Go-live:
  • Domain Configuration, SSL, Monitoring Dashboard, UAT and project defense.

6. Budget Estimation

Based on detailed cost estimates for the MVP phase.

You can review the detailed cost estimation by downloading the following files: 📊 CSV Pricing File 💾 JSON Pricing File

Other Costs:

(Based on On-Demand pricing in Singapore region - ap-southeast-1)

7. Risk Assessment

• Main Risks: User information leakage (Impact: High), AWS Region connection loss (Impact: High), AI misrecognition (Impact: Medium).
• Mitigation Strategies:
  • Security: AES-256 encryption, HTTPS, IAM Least Privilege, AWS WAF.
  • High Availability: Multi-AZ deployment for ECS and ALB.
  • AI: Continuously improve model with real data.
  • Resilience: Use internal RabbitMQ for asynchronous processing and retry.
• Disaster Recovery Plan: Use IaC (Infrastructure as Code) for rapid infrastructure restoration.

8. Expected Results & Team

Expected Results of the Project

• Automated financial data entry: The application helps users avoid manual entry, just take a photo of the invoice or record a voice for the system to automatically classify spending.
• Intuitive financial management: Users can view spending charts, monthly reports, and receive savings suggestions based on consumer behavior.
• Minimal user experience: Friendly web interface, modern design, optimized for mobile devices and suitable for people new to financial management.
• Stable, scalable system: Microservices architecture makes it easy to add new features such as spending reminders, AI predictive analysis, or expand to a mobile app.
• Improving development team skills: Project members have practical access to DevOps processes, CI/CD implementation, and cloud-based application optimization.

Project Limitations

• Vietnamese AI model is still limited: The ability to recognize regional voices or handwritten invoices has not yet achieved high accuracy.
• No separate mobile application: The MVP version only supports the web platform, there is no native mobile app.

Implementation Team:

Mentor Support:

• Nguyen Gia Hung - Head of Solution Architects
• Van Hoang Kha - Cloud Security Engineer