purujit goyal | Founding ML Engineer

Purujit Goyal Founding ML Engineer | Agentic AI | RAG | Streaming Systems India (IN) 487 connections • 489 followers About I'm a machine learning engineer focused on Agentic AI: multi-agent orchestration, RAG pipelines, real-time streaming systems, and LLM infrastructure. Most recently as a founding engineer, I built the end-to-end AI layer for a construction management platform: a LangGraph-based supervisor/worker agent system, a knowledge base RAG pipeline with contextual retrieval, and a CDC-driven universal search. Before that, I built real-time audio processing systems at Birch AI, handling ~10,000 calls/week, and worked on event-driven data pipelines at Rivigo using Kafka Streams. I've also done NLP research (NER, summarization with RLHF) and contributed to OpenReview's paper-reviewer matching system used by NeurIPS and ICML. I enjoy working across the stack: system architecture, ML research, and backend engineering, building impactful AI products. Published at NeurIPS, EMNLP, and ACL. Total Experience: 9 years and 2 months Experience Founding ML Engineer - Merlin AI Software (YC S24)(https://www.linkedin.com/company/merlin-ai-software) (Current) Jul 2025 - Present (1 year and 1 month) in Bengaluru, Karnataka, India Founding engineer for the AI layer of the construction management platform. Technical owner of the AI codebase. - Designed and built a production multi-agent system on LangGraph: a supervisor that routes user queries to domain-specific agents with tool use, agent-to-agent handoffs, and real-time streaming responses. - Built the Knowledge Base RAG system end-to-end: document ingestion and parsing, contextual embeddings into a vector database, and a search service used by both the AI agents and the product's Cmd+K palette. - Built a universal search spanning, combining keyword and semantic search with reranking, fed by a Kafka CDC pipeline that streams MongoDB updates into the vector store in real time. - Researched and develo