I started building AI tools on nights and weekends while working my day job as an inventory manager at a major film studio, teaching myself entirely by working alongside large language models. I had never written a line of code in my life, but I was so taken by the fact that I could learn anything I wanted just by asking the right questions that I fell in love with the process. Within two weeks, I was shipping things a junior developer would be proud of.
I started off with a RAG bot, then I made a Discord bot with an API key that was fine-tuned off of five years of my friend's messages in a Discord server. The idea that you could build a copy of your friend and put him in Discord, learn anything, build anything, just by asking the right questions. That hit me hard. It still does.
I grew up around people who built things with their hands and their vision. I think that's always been in me somewhere. I was always more of a fan of technology and realized I was more of a gearhead than a creative. AI was the combination of both. It gave me the medium that let me execute with something that also fascinated me.
When I built a working AI onboarding tool for my team at work, it got shut down by the corporate machine. Security concerns, internal politics, the usual. I stopped for a while. I was annoyed with the corporate way of doing things and felt discouraged. Eventually, a friend at work told me to check out Claude Code, and shortly after, OpenClaw. I was instantly hooked again. That decision led to Antaris Analytics LLC, and eventually to what I think is one of the more interesting infrastructure plays in the agent space right now.
Here's the part I get asked about most: I never wrote the code myself. Every line was written by my own AI agent, directed by me. I designed the architecture, made every product decision, defined the constraints, and set the direction. The agent executed. That's not a workaround. That's the proof of concept. If my own agent can build production-grade infrastructure that other developers are shipping with, that tells you something about what we've built.
Today the Antaris CORE is six production packages: parsica-memory, router, guard, context, pipeline, and contracts. Zero external dependencies. The zero-dependency principle wasn't a marketing decision. It was born out of necessity one night when a dependency crashed everything, and I decided to never let that happen again. We distill neural model knowledge into classical statistical structures, things like BM25 search and PPMI co-occurrence indexing, but add on things like word enrichment from whatever model you're currently running, so the system runs fast without dragging in someone else's supply chain. Benchmarks back up our hard work: 78% on the LOCOMO Benchmark, 88.5% across the board with doc2query, all while remaining zero dependencies. Every feature came from a real problem we hit in production, and instead of looking for solutions to download, we built everything ourselves.
Alongside Core, I built Parsica, currently a pre-alpha document search product powered by BM25+ with multi-format ingestion and a full domain pack library. Different product, same philosophy: keep it fast, keep it independent, solve real retrieval problems, maintaining 100% provenance, without bolting on a vector database and hoping for the best.
Then there's Antaris, a full-stack AI agent with Parsica-Memory built in, a 5-layer security environmental system, multi-provider routing with automatic failover, and a self-repairing runtime with a daemon watchdog. It learns behavioral patterns per user, tracks individual people to build personalized context over time, and adapts on its own. It plugs directly into the OpenClaw ecosystem through MCP, and you can migrate from OpenClaw or other agent and memory systems with a single command - copying your agent to the new body and leaving your old one intact. This is where everything in the stack comes together. The Core provides the infrastructure, Parsica handles retrieval, and the bot is the agent layer that sits on top of all of it and actually does the work.
Then an opportunity came my way with a very close friend of mine, and we partnered up for a week and began launching ForgeAI. He wanted to put part of my infrastructure on his bots, so we just built the entire weekend. That partnership is live now. Forge ships all three tiers with Parsica under the hood. Having my AI agents like Moro make me feel like I have a superpower, and it brings out this creative, engineering mindset that caters to what I do best: quick creative problem-solving and critical thinking. Building as much as we did in one week was unreal.
It's a surreal feeling to realize that there is a tool you can use to multiply your output by that much. I want to be a part of this, and that's why I started this company. I'm making a career pivot I probably should have made sooner. This is the most engaged I have ever been by any work in my life. Antaris is where I'm putting everything.