Case Study
intercoach
Real-time AI mock interview app with voice conversations, guided question flow, and instant feedback.
Context
Interview practice tools are often expensive, static, or disconnected from realistic voice-based conversation flow.
Approach
Built an open-source coaching app that runs local-first with pluggable LLM/STT/TTS providers and a guided interview loop.
Architecture
React + Vite frontend -> interview engine (SSE streaming) -> provider adapters (Ollama/OpenAI/Groq + Kokoro + Web Speech API) -> voice playback and avatar lip-sync.
Results
Delivers real-time mock interviews, instant feedback, and natural voice interaction in one workflow designed for repeated practice.
Lessons
Provider abstraction and local-first defaults reduce setup friction while keeping the system flexible for cloud and hybrid deployments.
Artifacts
If I had 2 more weeks...
Add longitudinal performance tracking, role-specific rubric packs, and deeper behavioral answer scoring.