Archit Gupta

Archit Gupta I'm Archit Gupta, a Founding Engineer building multi-agent runtimes, memory/RAG infrastructure, and production agent systems. Previously AI Systems Engineer at Robynn AI. I write about AI engineering at architg.me/blog I build agentic AI systems that have to survive real constraints, not just demos. At Robynn AI, I work on the backend and product surface of a production AI platform built around a SvelteKit application, a Python/LangGraph backend, a multi-agent registry, 60+ tools, and deployed LangGraph workflows for multi-step reasoning and execution. My work spans typed MCP-style workflows, Graphlit-backed memory and context infrastructure, long-running conversational runtimes, and observability for runs, tools, latency, and cost. I also redesigned a multi-LLM reporting pipeline from direct HTML generation to structured JSON + templates, cutting runtime from roughly 320 seconds to 46 seconds while improving reliability and reducing cost. Outside of work, I build systems like Midas, a four-agent continuous learning stock market-intelligence system that uses structured handoffs instead of shared memory; Ronnie, a multimodal pool-strategy agent grounded in a large billiards-specific

Career