Priyanshu Pratap Singh

I’m a Backend & AI Engineer building production-grade multi-agent systems, RAG pipelines, and LLM orchestration platforms. I’ve shipped 0→1 AI infrastructure on AWS with real-time streaming, MCP integrations, and Human-in-the-Loop workflows.

At SVAHNAR, I engineered a LangGraph multi-agent orchestration engine with PostgreSQL-persisted agent state, Celery async task queuing, and FastAPI SSE token streaming. I integrated MCP servers (STDIO & HTTP) to attach custom tools to agents and implemented HITL pause/inspect/edit/reject flows to reduce hallucination risk in production.

I’m also building ARCA, an open-source AI agent runtime with semantic memory that improves output quality across runs without fine-tuning. Through projects like Voice with RAG, I’ve designed agent routing, real-time WebSocket token streaming, and end-to-end RAG using BGE-M3 embeddings, pgvector/Qdrant, and live web retrieval with Tavily—optimized for real UX with interrupt/resume.