Skip to content
View Yog-Sotho's full-sized avatar
πŸ’­
Chillin' and vibe-coding πŸ€–
πŸ’­
Chillin' and vibe-coding πŸ€–

Block or report Yog-Sotho

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Yog-Sotho/README.md

🜏 Yog-Sotho

I build smart machines. Then I poke them to see what breaks. Then I fix it. Repeat until 3 a.m.

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

Profile views Telegram: @yogsotho GitHub followers


πŸ€” What is this guy up to?

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
  • πŸ›‘οΈ 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
  • πŸ§ͺ Experimenting in public: some projects ship, some explode, all of them teach me something

βš—οΈ The Lab (featured projects)

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 β€” grounded, judged, deduped. No code. Free to run, tested nightly.
🧠 LLM-fine-tuner Your own custom AI with no coding, no PhD and no drama. Upload data, click Train, get a model you can run in Ollama or ship to the Hub.
πŸ›‘οΈ LiSa-LLM Because your cat-sized model deserves a lion-sized sandbox. Drop in a GGUF, get an OpenAI-compatible API, and the model runs in a padded cell. Signed releases, sandbox on by default.
🧼 LLM-Sanitizer-Pro Scrubs the junk and PII out of training data before it ever reaches a model. Streaming, multi-format, redaction that actually checks the Luhn digit.
πŸ”Ž Ollama-Security-Auditor Checks whether your Ollama setup is locked down or wide open to the whole internet.
πŸ” Homomorphic-Encryption-Service Computing on data it can't even read. Yes, that is a real thing.
πŸŒ‘ Nocturna-Trading-System Experimental ML crypto trading, built to dodge race conditions, blocking I/O and losing its mind in a crash.
πŸ•΅οΈ Wash-Trade-Scanner Sniffs out wash-trading patterns in market data.

🧭 My rules of the lab

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

βš™οΈ Toolbox

Python JavaScript PyTorch Hugging Face Gradio Streamlit Docker GitHub Actions Linux


πŸ“ˆ Proof of life

GitHub streak

Yog-Sotho's contributions over the last year


🌐 Come say hi


🜏 Support

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

  • πŸͺ™ BTC (BSC network): 0x4b038d095ccb5bd088b8cdc33d4a202372647f2c
  • πŸ’Ž ETH (ETH network): 0x4b038d095ccb5bd088b8cdc33d4a202372647f2c
  • 🚧 BNB (BSC network): 0x4b038d095ccb5bd088b8cdc33d4a202372647f2c

🜏 Ph'nglui mglw'nafh. Keep experimenting.

Pinned Loading

  1. LLM-fine-tuner LLM-fine-tuner Public

    No-code LLM fine-tuning: upload your data β†’ click Train β†’ get a model you can run anywhere. Gradio UI + CLI Β· SFT/DPO/ORPO/KTO/GRPO/reward models Β· distillation, vision & embedding models Β· Unsloth…

    Python 38 5

  2. Brainbrew Brainbrew Public

    No-code synthetic training data from your documents: drop in PDFs/TXT β†’ click Generate β†’ get grounded Q&A datasets. LLM judge Β· MinHash dedup Β· benchmark decontamination Β· PII scrubbing Β· Alpaca/Sh…

    Python 8 1

  3. LiSa-LLM LiSa-LLM Public

    Sandboxed, OpenAI-compatible LLM server for GGUF models: drop in a model β†’ get a ChatGPT-style API β†’ the model runs in a padded cell. llama.cpp engine Β· continuous batching + prompt cache (faster t…

    C++ 1

  4. LLM-Sanitizer-Pro LLM-Sanitizer-Pro Public

    Clean, de-identify and curate LLM training data at scale: PII & secrets redaction (regex + GLiNER NER), exact/fuzzy/semantic dedup, language ID, FineWeb-Edu/DCLM quality scoring, Gopher/C4/FineWeb …

    Python 1

  5. Nocturna-Trading-System Nocturna-Trading-System Public

    Self-hosted paper-trading robot for US ETFs on Alpaca: connect a paper account β†’ it trades a 60/40 behind hard risk limits β†’ watch it from your phone. Kill switch Β· volatility target Β· idle cash in…

    Python 6

  6. Fortinet-Hunter-2026 Fortinet-Hunter-2026 Public

    State-of-the-Art FortiOS Attack Surface Mapper- Fortinet Hunter is an offensive security research tool for Fortinet appliances.

    1