Yog-Sotho Yog-Sotho

🜏 Yog-Sotho

Mad scientist of LLMs & AI security.
No lab coat, lots of coffee, zero chill about broken things.

I live where AI meets security, which is a fun neighbourhood. Everything is on fire, but in an interesting way.

🧠 Teaching models new tricks: fine-tuning, evaluation, and keeping them on a leash
🛡️ Making the AI stack harder to break: sanitising what goes in and out, auditing local model servers
🧰 Building tools that normal humans can use: click a button, get a result, no PhD required
🧪 Experimenting in public: some projects ship, some explode, all of them teach me something

	Project	What it does
🍺	Brainbrew	Mad scientist + coffee machine for data. Dump in your boring PDFs, press one button, get god-tier synthetic training data. Free to run, tested nightly.
🧠	LLM-fine-tuner	Your own custom AI with no coding, no PhD and no drama. Upload data, click Train, done.
🧼	LLM-Sanitizer-Pro	Scrubs what goes into and comes out of language models, so they behave.
🔎	Ollama-Security-Auditor	Checks whether your Ollama setup is locked down or wide open.
🔐	Homomorphic-Encryption-Service	Computing on data it can't even read. Yes, that is a real thing.
🌑	Nocturna-Trading-System	Experimental ML crypto trading, built to dodge race conditions, blocking I/O and losing its mind in a crash.
🕵️	Wash-Trade-Scanner	Sniffs out wash-trading patterns in market data.

📏 Measure honestly. A claim needs a benchmark, and a benchmark needs its caveats written down.
🔒 Secure by default. Secrets stay on the server, things are private until you say otherwise.
✂️ Keep it small. Typed code, tests that mean something, docs a stranger can follow.
💸 Free first. If you need a credit card to try it, I go looking for a free way.

If you like my work, feel free to send me a tip:

_{🜏 Ph'nglui mglw'nafh. Keep experimenting.}