Sahan Harshajith – Software Developer & Data Science Enthusiast

TeaBlendAI is an industry-focused digital platform designed to modernize the Sri Lankan tea auction and trading process by combining online tea auctions, business analytics, AI-powered insights, and conversational automation in one system. The platform enables tea producers, sellers, buyers, and administrators to manage and analyze tea auction activities digitally. Users can work with auction lots, bidding, purchases, tea grades, suppliers, blends, and buyer activity while gaining insights through interactive analytics dashboards. A key part of the platform is its AI-powered chatbot, which allows users to interact with the system using natural language. Instead of manually searching through different pages and reports, users can ask questions about auction and business data, obtain tea-industry information, generate visualizations, and—where authorized—perform auction-related operations through conversation.

An automated, enterprise-grade Foreign Exchange Rate Analytics platform built on a Medallion Data Warehouse Architecture using Microsoft SQL Server, Apache Spark, dbt, Apache Airflow, Scikit-Learn, FastAPI, and Next.js. The platform ingests live spot exchange rates from dual independent APIs into a Bronze raw audit layer, transforms and flattens heterogeneous JSON schemas via PySpark into a standardized Silver layer with idempotent T-SQL MERGE upserts, and generates business-ready Gold analytics marts for rolling volatility, Month-over-Month (MoM) performance, and Isolation Forest ML anomaly detection.

A learning-focused end-to-end SQL Data Warehouse project built with Microsoft SQL Server to explore data engineering, ETL, data transformation, dimensional modeling, and business intelligence concepts. The project integrates multi-source CRM and ERP datasets using a Medallion Architecture (Bronze → Silver → Gold). Raw CSV data is ingested and preserved in the Bronze layer, cleansed and standardized in the Silver layer, and transformed into a business-friendly Star Schema in the Gold layer for analytical reporting. The project also includes exploratory analysis of sales, customers, products, trends, segmentation, and RFM metrics using advanced SQL.

A full-stack e-commerce platform developed for Kings Bear, a premium handmade teddy bear brand. The application provides a complete online shopping experience including product browsing, cart management, wishlist functionality, secure authentication, checkout, order tracking, and a dedicated admin dashboard for managing products, users, and orders. The platform is built with a Next.js and React frontend backed by a FastAPI REST API. Firebase Firestore is used for data persistence and Firebase Authentication handles customer authentication, while administrators use a separate protected authentication flow. Product images are managed through Cloudinary, and automated email notifications are sent for order confirmations and password reset requests.

A full-stack weather analytics application built to collect, process, and visualize real-time weather data for 30 global cities using OpenWeatherMap API. The application combines a Next.js frontend with a FastAPI backend and provides secure authentication, ranked weather analytics, detailed city-level metrics, and a scientific Comfort Index Score based on temperature, humidity, wind speed, cloudiness, and visibility. It also uses an in-memory caching system and background refresh process to improve performance and reduce unnecessary API requests.

A full-stack MERN e-commerce application developed as a university group project for the Web Technologies module. The system was designed for an online bakery and confectionery store, providing a customer storefront, admin dashboard, and RESTful backend API. The platform allows customers to browse and filter products, manage their shopping cart, place orders, and track order status. Administrators can manage products, upload product images, and monitor and update customer orders through a dedicated dashboard.

A modern full-stack personal portfolio website designed to showcase my software development projects, technical skills, certifications, education, and professional experience through an interactive and responsive user interface. The portfolio is built with Next.js 16, React 19, TypeScript, and Tailwind CSS, with Supabase PostgreSQL powering dynamic content such as certifications, competitions, and skills. It features a dedicated dynamic project showcase system, responsive desktop and mobile navigation, animated UI interactions, project galleries, a contact form powered by Resend, and a dedicated CV download API.