Siddharth | Generative AI + ML Engineer

I build production-grade, real-time voice agents, RAG systems, and multimodal AI—blending deep learning, systems programming, and deployment to ship reliable products.

Expertise

• Voice AI: low-latency STT/TTS with Faster-Whisper, VAD, RealtimeTTS (Coqui, Kokoro, Orpheus)
• Conversational agents: FastAPI + WebSockets, Redis, LangChain
• RAG: ChromaDB, Pinecone, HuggingFace LLMs
• Multimodal agents: audio • text • vision
• Deployment: Docker, Kubernetes, FastAPI backends, Streamlit/JS frontends
• Data: PostgreSQL, Redis; async pipelines, session/state management

Projects

• Real-time Voice Agent (ChatGPT Voice Mini)

  • Full-duplex WebSocket voice, partial/final transcripts, VAD, memory, entity extraction
  • Stack: FastAPI, Redis, PostgreSQL, JS | STT: Faster-Whisper | TTS: RealtimeTTS (Coqui, Orpheus)
  • GitHub: https://github.com/Siddharth0207/voice-agent-llama

• RAG-Driven QA Engine

  • Context-aware Q&A with Chroma/Pinecone + HuggingFace LLMs
  • Packaged for Streamlit/Flask

Tech

Python, JavaScript, SQL • FastAPI, LangChain, HuggingFace, PyTorch, WebRTC • Redis, PostgreSQL • Docker, Kubernetes • Whisper/Faster-Whisper, VAD, RealtimeTTS • Git, VS Code

Goals

• Modular, composable agent architectures
• Long-term memory, multi-turn reasoning
• Reproducible, well-documented releases
• Collaborations on voice-first AI UX

If you’re building real-time agents or production ML infra, let’s connect.

“Build systems that listen, speak, and understand — in real time.”