I'm Archit Gupta, a Founding Engineer building multi-agent runtimes, memory/RAG infrastructure, and production agent systems. Currently at a stealth startup; previously AI Systems Engineer at Robynn AI, where I shipped orchestration APIs, structured-output pipelines, long-running runtimes, and observability.
Founding engineer at a stealth startup—production agentic systems and multi-agent infrastructure from the ground up
A four-agent continuous learning stock market-intelligence system that runs daily across hundreds of equities linked to dozens of commodity categories, translating macro and commodity shifts into trade-oriented research and signals.
A fine-tuned multimodal pool-strategy agent that reconstructs table layouts from photos, grounds decisions in a large billiards-specific RAG knowledge base, and recommends shot sequences tailored to player ability, style, and evolving preferences.
A compact Node.js priority-queue proxy for a self-hosted 16-agent system on a single GPU, built to keep embeddings, memory, search, and RAG pipelines responsive under real load.
League-management software for competitive billiards with real-time match tracking, ELO-based ratings across 50k+ historical matches, cryptocurrency payouts for 600+ active players, and operational tooling for league workflows.
Built an LLM-based hiring workflow that parsed resumes, generated role-specific technical questions, and automated first-pass screening; won the LabLab AI hackathon and earned incubator interest.
Notes on multi-agent systems, production AI, engineering tradeoffs, and the odd pool lesson.
Running a fleet of coding agents under a written constitution—authorizations, case law, and a sovereign bound by the same rules.
Sixteen agents, one GPU: how a small Node.js priority queue kept embeddings and RAG responsive under load.
Midas architecture notes: specialized agents for commodity-linked equity research instead of one overloaded generalist.
Production patterns and pitfalls from building agents at Robynn AI and in operational automation.
Bleeding-edge agents on boring infrastructure: PostgreSQL, Redis, and why complexity budget matters.
Competitive pool as a systems lesson: setup for the next shot beats one flashy make.
Career notes on range across frontend, backend, infra, and AI—and why “what’s your thing?” is the wrong question.
Discipline over motivation: building pool and engineering skill by making practice non-negotiable.
I'm a Founding Engineer based in Boulder, Colorado, focused on multi-agent runtimes, memory/RAG systems, and production agent infrastructure. I'm currently at a stealth startup; previously I was an AI Systems Engineer at Robynn AI, where I shipped orchestration APIs, structured-output pipelines, long-running runtimes, and observability.
My background spans enterprise engineering at Cisco, PayPal, and Riverbed, plus independent contract work building hiring workflows, operational systems, and billiards-league software. I studied Computer Science and Astronomy at the University of Illinois Urbana-Champaign.
During my time at UIUC, I worked on cybersecurity infrastructure, configuring ExaBGP-based blocking and prototyping Google Rapid Response integration for incident response workflows.
Outside of tech, I play pool competitively, build AI systems around the game, live-stream matches on my YouTube channel, and perform Indian classical music on the Sarod.
Have a project in mind or want to chat about AI? I'd love to hear from you.
