Electron-Based Live Interview Application

Overview

Developed an Electron application that enables students and candidates of a client to participate in live interviews on both Windows and macOS. The product combined real-time audio capture, LLM-powered assistance, over-the-air updates, and platform-native integrations to deliver a reliable interview experience while meeting strict client and investor demo requirements.

Tech stack

• Electron
• TypeScript
• Node.js
• Objective-C
• AssemblyAI
• OpenAI
• Firestore
• Bugsnag
• RAG
• macOS Core Audio

Features implemented

• Cross-platform desktop support for Windows and macOS
• Over-the-air application updates with managed versioning
• Bugsnag integration for production error monitoring
• Automated initial configuration via AppleScript
• Custom Objective-C bridge into the Electron Node.js environment
• Silence detection to switch between speaker mic and system mic, preventing audio overlap
• AssemblyAI real-time transcription with ~2 second delay
• Automatic LLM-based question generation triggered by detected silence
• Resume analysis to extract candidate information
• RAG pipeline for personalized interview answers
• Magic link authentication using Firestore
• Fine-tuned OpenAI models for improved answer quality
• Real-time meeting audio recording with server-side transcription and post-meeting AI analysis
• In-app sharing of meeting transcripts and analysis
• Hidden screen recording feature to avoid detection during interviews
• macOS and Windows application notarization

Key achievements / impact

• Led the team and handled client communication four days a week, including sprint planning
• Presented product demos to clients and investors
• Maintained high code quality with TypeScript and established engineering best practices
• Delivered a production-ready dual-platform Electron product with native audio integrations