I'm not an ML engineer, secops specialist, or sysadmin. I'm someone who's spent the last several years living alongside AI agents — building with them, breaking them, routing between them, and figuring out what they're actually good for. I want to use this experience to help those who aren't quite as tech-fluent.
There are people who understand transformers at the math level. I'm not one of them, and I'm not going to pretend to be. What I am is someone who picked up AI tools early and never put them down. I've spent the last several years figuring out — through sheer hours of hands-on use — what these systems can do, where they fail, and how to make the most of them.
I've built agent systems that automate all sorts of things for me. I've tuned agents to write better code. I've routed requests across a dozen different providers to find the right tool for the job. I've typed until my usage limits were reached, every day. And then I spent more than I could afford on tokens via API.
What I care about now is where this goes. AI is reshaping how work gets done, and the people closest to the tools — not just the people building them — are going to be the ones who figure out how to make it actually useful in real settings. That's what I want to do. Not research. Not model design. Application. Putting working AI into the hands of people who need it. And yes, Anthropic, I'm looking at you.
I don't have a PhD. I have thousands of hours of actually using these tools until I understood them from the inside.
I've used every major model - Claude, GPT, Gemini, Open-soure variants (sorry, dario) - enough to know their individual personalities, failure modes, and sweet spots. I saw openclaw blow up, but I got on the hermes agent train very early. I know how to use AI more effectively than the vast majority of people.
I've used AI agents to build AI agent systems with tool-use, persistent memory, and multi-step reasoning loops. Designed orchestration that routes tasks, handles failures gracefully, and actually completes tasks, not just demos. Shipped to prod for personal use.
I originate from Silicon Valley, and both my parents were involved in the tech industry before the internet existed. I can type better than I can write (120wpm), I wrote my first website when I was 13 (it wasn't very good), I get agitated if I don't have a tiling window manager on my personal computer. I am and have always been the go-to tech teacher/debugger/antiviral individual for the normies around me.
I've connected AI models to all sorts of stuff — graph retrieval, tool-use, MCP servers, browser automation. I'm currently mastering the plumbing that turns a model endpoint into something that does a job.
Built and daily-use a multi-agent system that automates real work — research, code generation, file management, web interaction. Tool-use loops, persistent memory across sessions, dynamic model routing across providers. It's not a demo. It runs my life.
Spent months probing how different models behave under stress — adversarial inputs, edge-case prompts, capability boundaries. Not academic research, just my borderline autistic need to know how things work. Building intuition on how models break and why.
Built pipelines that connect AI models to real-world systems — scraping, processing, and acting on data autonomously. Integrated multiple model providers with fallback logic. Made AI do things that used to take humans hours.
Set up local inference pipelines on Apple Silicon — running quantized models, building custom serving setups, fine-tuning for specific tasks. Learned the practical tradeoffs between model size, speed, and quality by running hundreds of experiments.
Both pro and personal work as a repair tech — hardware upgrades, diagnosing issues, removing malware, usage lessons. That background shaped how I think about AI: I look for ways to abstract away the technical stuff so a normie can achieve the same results someone with tech literacy can.
I'm applying to Claude Corps because I believe the people who will make AI useful in the real world aren't just the ones building the models — they're the ones who've been using them long enough to know what works. I've got the hours. I've got the obsession. I want to put it somewhere it matters.