Marketing website for Biz Guide, a Sri Lankan company registration, secretarial, bookkeeping, payroll and tax firm.

What I delivered: • 4-page site: Home, Company Registration, Services, Contact • Enquiry form that sends by email or WhatsApp, with a form-backend option • Built-in SEO: structured data, sitemap, Open Graph tags • Automatic image optimisation (AVIF/WebP)

Lighthouse (mobile): 100 Performance, 100 Accessibility, 100 Best Practices, 100 SEO.

Tech: Astro, Tailwind CSS, TypeScript

A concept redesign for a Japanese-style cat café in Colombo, made as a pitch to the owner.

The page sets a cosy, hand-made mood with Ghibli-inspired visuals and smooth scroll animations. It also answers what visitors need: opening hours, location, menu and café rules. Main actions are "Get directions" and "Call".

Sections: story, resident cats, full menu, a drawing wall, and a visit/hours section with a live "open now" status.

Tech: HTML, CSS, JavaScript, GSAP, Lenis. No build step, so it deploys anywhere.

A learning platform for GCE O/L ICT students to learn Pascal by writing real code in the browser.

Features: • Custom Pascal interpreter built from scratch (lexer, parser, type checker) with friendly error messages • Code Lab with a step-by-step "show me what happened" visualiser • 24 topics, 10 practice question types, boss battles and mini-games • Mock exams with a readiness score • XP, streaks and 27 achievements • 331 automated tests

Tech: React, TypeScript, Vite

An interactive learning site covering all 12 levels of the A/L ICT data communication and networking syllabus.

Every waveform, topology and diagram is calculated live from real equations, with no static images. Students drag sliders and watch signals change.

Labs include signal properties, line encoding (NRZ, Manchester), modulation (ASK/FSK/PSK), multiplexing, network topologies and quizzes.

Tech: React, TypeScript, Vite, SVG

An interactive course for A/L ICT IoT and Arduino lessons, built for students with no lab access.

Every circuit in the syllabus is wired and running in the browser, and every program actually runs: change a number, press Run, and watch the LED respond.

Includes lessons, a free-play lab, practice, an exam mode and progress tracking.

Tech: React, TypeScript, Vite, custom Arduino interpreter

A real-time drawing and colouring app you control with hand gestures through a webcam.

AI parts: • MediaPipe hand tracking turns finger gestures into drawing tools • A U-Net edge-detection model trained from scratch on the BSDS500 dataset turns any photo into clean outlines you can colour in • Multi-scale inference, Canny fusion and gap bridging keep outlines closed and paintable

App features: shapes, 20 colours, eraser, undo/redo, zoom and pan, flood fill.

Tech: Python, PyTorch, OpenCV, MediaPipe

A machine learning pipeline that classifies skin images as psoriasis or normal.

Pipeline: • Image cleanup: hair removal, denoising and GrabCut segmentation • Deep features from ResNet-50 and EfficientNet-B0 combined into one 3,328-value vector • Random Forest, SVM and KNN combined with soft voting

Results on the test set: 91% accuracy and 0.963 ROC AUC.

Tech: Python, TensorFlow, scikit-learn, OpenCV

An interactive dashboard that predicts a film's box-office revenue from its budget, runtime, genre and release date. Built as a university team project.

• Ensemble model (Random Forest, XGBoost and LightGBM) trained on 13,000+ movies, 88.4% accuracy • ROI gauge, best-release-month chart and best-season advice • Scenario analysis comparing conservative, base and optimistic budgets

Tech: Python, scikit-learn, XGBoost, LightGBM, Streamlit, Plotly

A personal monogram logo and mini brand system.

Process: I explored six concept directions, then chose a geometric "SP" mark. Every curve uses one radius and one stroke weight, so the logo stays crisp from a 16px favicon up to print.

Delivered: • Dark, light, mono and colour versions • Horizontal and vertical lockups • App icons and social media avatars • A one-page brand guide

A live online store I designed and built from scratch for a collectibles business (figures, books and apparel).

Storefront: • Product catalogue with search, category and price filters • Product pages with image gallery, ratings and stock badges • Cart, wishlist and checkout with cash on delivery and bank transfer (with payment-slip upload) • Customer accounts, order history, reviews and newsletter • AI helper that explains products to new customers

Admin dashboard: • Manage products, stock alerts, homepage banners and menus • Process orders, review payment slips, discounts and custom orders

Also: SEO setup, rate limiting and automated tests.

Tech: Next.js, TypeScript, Supabase (PostgreSQL), Tailwind CSS, Zustand