About me

I’m Max. I’m a computer science student at York University in Toronto, and I build software end to end.

What I actually like doing is AI. I’ve spent a lot of time deep in it: using models every day, pulling them apart to see how they work, and figuring out how to make them do something useful. It’s the part of the field I can’t put down.

Where I want to end up is research. I want to work on powerful models themselves, build them, modify them, and find the best way to use them as agents. My bet is graphs: models wired together so they plan, check their own work, and build real systems instead of one model guessing at one prompt. That thread runs through everything I’m making right now.

I care about open source too. The interesting work is happening in the open, and I’d rather stack open models I can run and change than rent a black box I can only poke at.

Right now I’m building Sequence, a canvas that turns system architecture into a checkable graph, and Project Brain, an internal knowledge base at SCHWAI. That second one is real DevOps and MLOps work: standing up the backend, the ingestion workers, and the open-source models that read the company’s own data.

Before this I shipped production features at Neo Financial, a bank serving more than a million customers. That’s where I learned what “it works on my machine” costs when the machine is someone’s actual money.

Most of my time goes to Sequence right now. It’s a canvas where you draw a system’s architecture as a graph, that graph renders a deterministic brief an agent builds the first version from, and a CI check fails the build the moment the code drifts from the design. I’ve written it up in the deep dives.

At SCHWAI I’m building Project Brain, the company’s internal knowledge base. It’s DevOps and MLOps work end to end: the backend, the ingestion workers, and open-source models reading the company’s own corpus. It gets more useful the more people lean on it, which is the whole idea.

The bigger goal underneath all of this is research. I want to work on the models themselves, build them, modify them, and figure out the best way to use them as agents. My current answer to that last part is graphs: feed an agent structured, deterministic data and it builds the right thing instead of guessing.

I lean open-source wherever I can this year. Models I can run, read, and change beat an API I can only call.

And I’m still a CS student at York, with school as the steady background process behind all of it.

The stuff I actually use, not a wishlist.

Languages are Python and TypeScript, mostly. Some of everything else when a project needs it.

On the backend I reach for FastAPI or Node with Express, Postgres as the default database (usually through Supabase), and Redis when something needs to be fast or queued. On the frontend it’s React, Next.js when the project wants a framework, and Tailwind for styling. Docker keeps it all reproducible.

When something in my workflow gets repetitive, I’d rather build my own tool for it than wire up an off-the-shelf automation platform. A tool I wrote fits my exact process, and I understand every line, so when it breaks I know why. Usually that’s a small script or a CLI, not a whole app.

For deploys, Vercel for anything frontend-shaped, Render when I need a real server, Cloudflare for the edge and DNS.

Nothing exotic here. Boring, well-understood tools let me spend the interesting hours on the actual problem.