Interview Copilot is a premium, mobile-first Progressive Web Application (PWA) designed to assist candidates with real-time feedback during technical interviews. Operating entirely offline-first, it captures and processes audio locally to guarantee total privacy, low latency, and zero cloud server dependencies.

Key Features

• Real-Time Voice Clarify & VAD Overrides: Mobile-optimized touch HUD targets supporting both "Hold-to-Talk" gestures and "Single-Tap Toggle" overlays for seamless hands-free transcription control.
• Dual-Input System/Speaker Capture: Easily captures local Microphone input or toggles to record system speaker audio output (e.g., Zoom, Google Meet, or browser audio streams) natively via the Screen Capture API.
• Multi-JSON Q&A Database Imports: Dynamically loads, parses, and merges concurrent custom Q&A .json question libraries on the client side, avoiding database locks.
• Background-Threaded Vector Search: Semantic search powered by a Web Worker that manages high-dimensional vector math and returns semantic matches instantly.

Technical Architecture & Role

As the Lead Software Architect for this project, I developed the PWA's architecture focusing on extreme offline efficiency and main-thread responsiveness:

• Frontend & UX: Next.js App Router, styled with high-end glassmorphism HUD design tokens, fluid dark-mode theme HSL values, and 60 FPS micro-animations.
• Offline Database Engine: SQLite WASM compilation initialized client-side in background Web Workers. Queries utilize SQL transactions and customized indexing techniques, speeding up loads by up to 50x.
• Background Vector Core: Offloaded heavy high-dimensional cosine similarity matching and FileReader matrix-merging algorithms to standalone Web Workers, maintaining exactly zero main-thread block time.
• Resilient Offline PWA Capability: Full offline capabilities enabled through advanced Service Worker strategies, pre-caching web assembly packages (sql-wasm.wasm), fonts, and bundle chunks for instant startup.

Skills and Deliverables

React & Next.js Web Workers SQLite WASM Vector Similarity (RAG) PWA & Service Workers Vitest & Playwright E2E