Description
Co-developed Owlyn, a real-time multimodal AI interview and assistant system that analyzes audio, video, and workspace signals during technical sessions. Designed a distributed architecture with an Electron client, backend orchestration layer, and multiple specialized agents for voice interaction, visual analysis, and code reasoning, enabling sub-second conversational feedback and synchronized live transcripts.
Highlights
- Co-developed Owlyn, a real-time multimodal AI interview and assistant system capable of analyzing audio, video, and workspace signals during technical sessions.
- Designed a distributed architecture with an Electron client, backend orchestration layer, and multiple worker agents handling voice interaction, visual analysis, and code reasoning.
- Implemented low-latency streaming pipelines using WebRTC and LiveKit, enabling sub-second conversational feedback and synchronized transcripts.
- Integrated Redis for live session state and PostgreSQL for persistent storage, ensuring consistency across concurrent agent interactions.
- Engineered a multi-agent system where specialized agents handle conversation, workspace analysis, and post-session evaluation.