Projects

A blazingly fast, lightweight, open-source native web browser for macOS built with Swift, SwiftUI, AppKit, and WebKit (WKWebView). Engineered for extreme memory efficiency (~150MB baseline RAM), zero telemetry, and a distraction-free Zen browsing experience.

• Engineered an intelligent tab-decay engine that automatically transitions idle tabs into deep sleep, freeing underlying WebContent processes while instantly restoring state on focus.
• Built native ad and tracker blocking with over 145,000 compiled rules from EasyList and EasyPrivacy converted into native WKContentRuleList bytecode with cosmetic DOM element hiding.
• Implemented unified WebKit process pooling (WKProcessPool), background tab concurrency limiting, and smart video defusal scriptlets.

A fluid, contextual Dynamic Notch and Island utility for macOS built natively with Swift, SwiftUI, and AppKit. Transforms the MacBook notch and external displays into an interactive desktop command center with spring physics, hardware telemetry, audio spectrum visualizer, shelf staging, and local AI agent monitoring.

• Built an adaptive notch and floating island interface conforming to physical Mac notches and external monitors with spring-damped expansion and frosted glassmorphism.
• Engineered a real-time hardware-accelerated audio FFT visualizer using Apple's Accelerate framework (vDSP) and system loopback audio capture.
• Integrated a drag-and-drop file shelf with a detachable floating desktop basket, visual clipboard manager with Apple Vision OCR, system HUD overlays, and live monitoring for local AI coding agents.

A minimal, native macOS edge-docked sticky notes utility built with Swift, SwiftUI, and AppKit. Designed to live subtly on the screen bezel as compact idle capsules that smoothly fan out with proximity awareness for instant scratchpads, checklists, and note capture without stealing window focus.

• Engineered a proximity-aware fan deck with custom spring-damped cascade physics, fanning out edge tabs on cursor approach and expanding notes without stealing system focus via non-activating NSPanel.
• Developed procedural scribble checklists featuring custom animated pen-drawn checks and organic strikethroughs powered by de Casteljau cubic bezier splines and FNV-1a seeded randomization.
• Built an atomic JSON persistence engine with zero telemetry, custom handwriting typography rendering (Chalkboard SE & Caveat), global hotkey toggles via Carbon Events, and Sparkle 2 automatic updates.

An offline-first emergency assistance mobile application tailored for low-literacy users and high-stress scenarios, featuring high-contrast pictorial guidance, step-by-step first-aid protocols, and dynamic CPR pacing.

• Built a native-language accessibility engine supporting 11 regional languages combining pre-recorded localized audio cues and Text-to-Speech voice narration.
• Developed a one-touch SOS dispatch system with a 3-second hold-to-trigger safety mechanism, real-time GPS coordinate geocoding, and automatic multi-recipient emergency SMS alerting via a custom native Android SMS module.
• Implemented an intuitive Symptom Checker and Emergency Triage flow with local caching, ensuring 100% offline access to critical life-saving guides without internet connectivity.

A full-stack collaborative note-sharing and social learning platform enabling users to create, organize, and share study materials through hierarchical folder structures and rich note management.

• Implemented real-time direct messaging with Socket.io, supporting live typing indicators, read receipts, and online or offline presence tracking.
• Developed a social timeline and notification system using Server-Sent Events with aggregated feeds based on user network relationships and activity.
• Created an interactive analytics dashboard using Recharts to visualize engagement metrics, note activity, and user interaction trends.
• Integrated secure authentication using bcrypt, media handling via Multer and Cloudinary, and optimized global state management with Redux Toolkit and Redux Persist.

A high-performance Retrieval-Augmented Generation system for indexing, analyzing, and querying software codebases in real time through Web UI, TUI, and CLI interfaces.

• Implemented AST-aware semantic parsing using Tree-sitter to extract functions, classes, and imports across TypeScript, JavaScript, Python, and Java codebases.
• Designed an incremental indexing pipeline using SHA-256 file hashing to avoid redundant reprocessing and automatically prune deleted or modified files.
• Developed repository-local vector storage using ChromaDB with isolated directories and process-safe concurrency control via file locking.
• Engineered a 1-hop call-graph expansion system to resolve cross-file dependencies and eliminate retrieval blind spots in multi-file projects.
• Built an interactive 2D force-directed codebase graph visualizer using React and react-force-graph-2d, and integrated NVIDIA NIM with Llama 3.1 8B Instruct for context-aware, streaming, token-budgeted responses.
• Provided multiple interfaces including a modern React dashboard, a terminal-based TUI, and a CLI tool for indexing and querying repositories.