Financial Audit RAG System with Chain-of-Thought
Status: Thesis project (2025-2026)
Problem: Financial audit requires analyzing hundreds to thousands of pages per report across multiple companies and years. Manual review is slow and inconsistent.
What I built: RAG system using Qwen-3 with Chain-of-Thought prompting to analyze financial and annual reports across 15 companies from different sectors, covering 4 years of data.
Tools: Python, Qwen-3 8B, Chain-of-Thought, RAG architecture, Vector Search
Key result: Implemented CoT prompting and compared against Qwen-3 8B baseline to improve response reliability and reduce hallucination.
GitHub: AuditCoT-RAG
Multi-Cloud Landing Zone for Banking Applications
Status: Architecture project (2025)
Problem: Banking applications need secure and compliant cloud environments with proper isolation and access controls.
What I built: Designed Landing Zone architectures for 2 enterprise clients across 3 OCI compartments and 5 AWS accounts with least-privilege IAM policies and Security Groups. Created documentation templates reused for future client deployments.
Tools: OCI, AWS, Terraform, CLI, IAM, Security Groups
GitHub: BankXYZ-AWS-LandingZone
Fintech Fraud Detection - RISTEK Datathon 2024
Status: Competition project (2024)
Competition: Top 100 out of 200 to 300 university teams.
Problem: Banking transaction dataset requiring an ML model to predict fraudulent transactions.
What I built: Developed ML model for fraud detection with comprehensive data preprocessing and feature engineering.
Tools: Python, Scikit-learn, Pandas, Jupyter Notebook
Brain Stroke Prediction using Random Forest
Status: Course and personal project
Problem: Predict patient likelihood of brain stroke based on demographic and health data such as age, BMI, glucose levels, and hypertension.
What I built: ML pipeline with EDA, preprocessing, categorical encoding, outlier handling, SMOTE balancing, and Random Forest classifier.
Tools: Python, Scikit-learn, Pandas, Matplotlib, Seaborn, SMOTE
Key result: Deployed Decision Tree and Random Forest classifiers with accuracy metrics, classification reports, and confusion matrices.
GitHub: StrokePredictionAnalysis
MedCare - Integrated Healthcare Management Platform
Status: Course project
Problem: Healthcare management is fragmented. Patients and doctors need a unified platform.
What I built: Comprehensive healthcare app for appointments, room reservation, medication purchase, medical history, hospital locator, emergency help center, and doctor-side room plus appointment management.
Tools: Full-stack web development
GitHub: MedCare
Data Warehouse for Automobile Manufacturing - ETL Pipeline
Status: Course project
Problem: Automobile manufacturing company needed a centralized data warehouse for strategic decision-making.
What I built: Designed a centralized data warehouse and implemented end-to-end ETL processes using Pentaho on 1,000 to 2,000 data sets.
Tools: Pentaho, SQL
GitHub: DataWarehouseCarManufact
Cardiovascular Disease Risk Prediction
Status: Course and personal project
Problem: Predict cardiovascular risk using clinical biomarkers.
What I built: Regression models with advanced feature engineering on 2,000 to 3,000 clinical records from Kaggle.
Tools: Python, Scikit-learn, Pandas
Key result: Achieved around 90 percent accuracy after optimization.
GitHub: HeartDiseaseModelTraining
Social Media Comment Username Scraper
Status: Full-stack web app (Flask)
Problem: Investigating account activity across social platforms can be time-consuming when done manually.
What I built: Flask app to search target usernames in comments and replies across Instagram, Facebook, YouTube, TikTok, and X, with per-platform input forms and a unified result summary page.
Tools: Python, Flask, Selenium, HTML, CSS, JavaScript
Key result: Added screenshot preview with modal zoom and pan, plus responsive UI for desktop and mobile workflows.
GitHub: social-media-username-finder
Hotel Management Desktop Application
Status: Desktop systems project
Problem: Hotel operations need integrated workflows across reception, administration, housekeeping, and inventory.
What I built: Windows Forms application with authentication, guest registration, payment and checkout flows, reservation support, employee and payroll handling, inventory, and housekeeping supply modules.
Tools: C#, .NET WinForms, SQL Server, Crystal Reports
GitHub: Hotel-Management-Application
AIRA Flight Booking Web Application
Status: Web application project
Problem: Airline booking systems need smooth end-to-end user flows from search to payment and invoice.
What I built: PHP and MySQL web app covering registration/login, flight search, ticket ordering for multiple passengers, payment processing, invoice generation, and profile updates.
Tools: PHP, MySQL, HTML, CSS, JavaScript
GitHub: aira-flight-booking
IRIS OCI Chat Platform (Recovered Full Stack App)
Status: Full-stack cloud-ready project
Problem: Needed a reconstructed chat platform with persistent history and cloud deployment readiness.
What I built: Recovered frontend plus Node.js backend and database model with guest mode, cookie-based auth, chat history persistence, and OCI deployment baseline with Terraform and Docker support.
Tools: Node.js, Express, SQLite, HTML, CSS, OCI Terraform, Docker
GitHub: iris-oci-chat-platform
Multi-Cloud Object Explorer Dashboard
Status: Cloud operations dashboard
Problem: Managing buckets and objects across multiple cloud providers is fragmented.
What I built: Unified dashboard for AWS S3, OCI Object Storage, and Cloudflare R2 with bucket explorer workflows and backend integration for cloud operations.
Tools: Node.js, HTML, CSS, JavaScript, AWS, OCI, Cloudflare R2, Terraform
GitHub: multicloud-object-explorer
Product Recognition System
Status: Machine learning project
Problem: Build a model capable of recognizing product categories from image data.
What I built: End-to-end Python-based product recognition workflow including dataset preparation, model experimentation, and evaluation.
Tools: Python, Machine Learning
GitHub: product-recog