Back to Projects Jun 2026

FinSight

Project Manager & Data Scientist

An AI-powered financial assistant web application designed to help MSMEs predict bankruptcy risk and strategize capital planning. Led a cross-functional team of 6 to build the platform integrating Laravel, FastAPI, a custom Keras Deep Learning model, and Gemini AI.

Laravel Python TensorFlow Gemini AI MySQL Tailwind

DOCUMENTATION

// ABOUT_PROJECT

Developed during the Dicoding Coding Camp 2026, this platform serves as a comprehensive AI-powered assistant designed to support Indonesian MSMEs with bankruptcy risk mitigation and strategic capital planning. The architecture centers on a hybrid system, combining a robust Laravel web application with a dedicated FastAPI inference service that hosts a custom Keras Deep Learning model. To ensure high predictive accuracy, the model incorporates a custom FinancialInteractionLayer to capture complex relationships between indicators and WeightedFocalLoss to address extreme class imbalance, achieving a 97.2% accuracy and 0.84 AUC-ROC. As the Project Manager and Data Scientist, I led the team to integrate these prediction endpoints, implement a Gemini-powered business advisor chatbot, and design an interactive Streamlit EDA dashboard. The final product bridges the gap between advanced financial intelligence and daily micro-business operations, empowering small business owners with instant, actionable risk scoring and capital planning.

MORE PROJECTS

Accenprove

A web-based digital reporting platform featuring a secure Role-Based Access Control (RBAC) system, part of Asah's program capstone project. Led a cross-functional team of 5 to develop the platform for digitizing official reports with automated document generation.

Billens

An AI-powered bill splitting web application that uses Agentic Vision to extract receipt data from photos and automatically calculates each person's fair share including proportional tax, service charges, and discounts. HKI-certified for its inventive computational logic.

BlinkSafe

An IoT-based wearable device awarded 3rd Winner in the PKM-KC competition at Udayana University, developed to mitigate road accidents through real-time physiological monitoring. Engineered a responsive detection system using integrated Infrared and Gyroscope sensors.