The hardest part of building Hoovik — my open-source AI-powered meeting platform — wasn’t WebRTC signaling or media pipelines.

It was managing real-time multimodal inference (PyTorch, MediaPipe, AudioWorklets) across distributed services without blocking the event loop or exhausting memory when packets drop, connections fluctuate, or webcams disappear.

🔗 https://dev.to/anupam_kumar/the-hardest-part-of-building-an-ai-powered-webrtc-platform-wasnt-webrtc-19bl

#WebRTC #AI #MachineLearning #OpenSource #NodeJS #FastAPI #PyTorch #EmotionAI #BuildInPublic #Python #Fediverse

Built Hoovik — an open-source WebRTC meeting app with real-time multimodal emotion AI using PyTorch, XGBoost, MediaPipe, Redis & FastAPI.

Deep technical breakdown of the architecture:
https://dev.to/anupam_kumar/inside-hoovik-building-a-real-time-multimodal-emotion-ai-pipeline-5267

Code & Demo:
- GitHub: https://github.com/AnupamKumar-1/Hoovik
- Live: https://hoovik.onrender.com

#OpenSource #WebRTC #MachineLearning #WebDev #Python #ReactJS #FastAPI #Redis #PyTorch #AI

Built Hoovik — a distributed real-time multimodal AI platform for live video meetings.

Features:
• WebRTC + Socket.IO signaling
• Real-time multimodal emotion inference
• FastAPI + PyTorch backend
• MediaPipe facial tracking
• XGBoost ensemble calibration
• IsolationForest anomaly detection
• Distributed Azure compute deployment

Live demo:
https://hoovik.onrender.com

Open source:
https://github.com/AnupamKumar-1/Hoovik

#Python #MachineLearning #MultimodalAI #WebRTC #OpenSource #FastAPI