I'm Thomas Kawas, founder of Meridian. I build production AI systems and ship them end to end. Right now that means building Meridian solo: a production, agentic GraphRAG platform, live on Vercel and Supabase.
Under the hood, Meridian is a lot of moving parts working as one: an agentic AI pipeline, knowledge-graph retrieval, an agent memory and context layer served through a custom MCP server, and an evaluation harness that validates the system before anything ships. For five years alongside this work I've been in customer-facing technical roles, currently leading Customer Success across a portfolio that runs from small teams to government, so I'm as comfortable explaining a complex system in plain language as I am building it. I'm ISC2 Certified in Cybersecurity, and I'm currently building a real-time autonomous agent that perceives from raw pixels with computer vision and learns human-like behavior through reinforcement learning.
I don't just build growing technology, I've lived it. I've grown plants with real success in controlled indoor environments: tent setups running automated self-watering systems and AI-driven sensors managing the climate, and I've gone deep on the science underneath it, from plant nutrition to disease to cultivation. Meridian is built by someone who actually understands the problem, not only the code.
The way I work is first-principles and relentless. I find a real problem, build the solution myself, and refine it until it’s genuinely right, not just “done.” Meridian’s agent context system went through that loop more times than I can count. I build to understand every layer of what I ship, which is how one person runs a system this complex in production. What drives me is the hard, unsolved edges: I’d rather sit with a genuinely difficult problem and find my own way through it than reach for a template. I build because I love the craft, and going deep on a problem until it clicks is the part I’d do whether or not anyone paid me for it.
And I build for the long game. Every decision is made to still hold years from now, at scale, because what I'm building isn't a tool, it's the plant atlas: a system that understands plants the way nothing before it has. Most people can't picture it yet, and that's expected. I can see exactly where it goes, and I'm building like it's already inevitable.