Cuanversation is a machine-learning–powered portal designed to help banking sales teams focus on higher-potential customers during deposit campaigns. By providing lead scores, customer tagging, historical insights, and guided calling tools, the platform aims to reduce wasted outbound calls and support more effective, data-driven sales operations.
| No | ID | Name | Learning Path |
|---|---|---|---|
| 1 | R269D5Y1236 | Muhammad Dony Mulya | React & Back-End with AI |
| 2 | R013D5Y1216 | Muhammad Aprilianto | React & Back-End with AI |
| 3 | R657D5Y0492 | Dimas Rizky Maulana | React & Back-End with AI |
| 4 | M117D5X0554 | Fadhilah Nurrahmayanti | Machine Learning |
| 5 | M891D5Y0020 | Abiyyu Rasyiq Muhadzzib | Machine Learning |
The ML pipeline predicts the probability that a customer will open a time-deposit product. It is built as a binary classification model that outputs a score between 0 and 1, which feeds into the backend scoring service.
Key steps:
- Exploratory Data Analysis (EDA)
- Data preprocessing
- Model selection
- Iterative training, evaluation, and parameter tuning
- Export to ONNX for deployment
The backend follows Clean Architecture, separating domain logic, use cases, repositories, delivery handlers, and supporting platform utilities.
It exposes a REST API and is designed for efficient querying and smooth interaction with campaign and lead data.
- Tech: Go (Fiber), PostgreSQL, RabbitMQ, Redis
- Infra: Azure, Docker, Prometheus, Grafana, Loki
- CI/CD: GitHub Actions
The frontend is built as a Single Page Application and follows an atomic design structure to keep components organized and reusable.
Users log in through a JWT-based system with role distinctions between managers and sales. The overall design and layout follow styles and components referenced from the project's Figma work.
- Tech: React, Vite, TypeScript, React Router, Zustand, Zod
- Styling: TailwindCSS, Ant Design, Framer Motion, Recharts
