Skip to content
View ademyanchuk's full-sized avatar
2 followers · 4 following
  • Hannover, Germany

Achievements

Achievement: Pull SharkAchievement: Arctic Code Vault Contributor

Achievements

Achievement: Pull SharkAchievement: Arctic Code Vault Contributor

Block or report ademyanchuk

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
ademyanchuk/README.md

Hi, I'm Alexey 👋

I'm a Machine Learning researcher currently working at the intersection of ML and medical research. Alongside my professional work, I'm independently deepening my technical skills by building large language models from scratch.

Currently:

  • 🎯 Working through Andrej Karpathy’s "Zero to Hero" series
  • 🛠️ Building and training a from-scratch GPT-2 model (gpt2-diy)
  • 🧠 Focusing on understanding transformers, scaling behaviors, and efficient training dynamics

I'm particularly interested in:

  • Large Language Model (LLM) internals
  • Model interpretability and robustness
  • Scalable and responsible AI development

Current Projects:

  • minbpe — A minimal, from-scratch BPE tokenizer
  • gpt2-diy — Reproducing GPT-2 from scratch as a learning journey

Building deep foundations, one token at a time.
Always happy to connect with fellow researchers, engineers, and builders in the ML/AI space!

📫 How to reach me: LinkedIn

Pinned Loading

  1. minbpe minbpe Public

    Educational reimplementation of Byte Pair Encoding (BPE) with regex tokenization, merge order control, and chunk-based encoding – inspired by Karpathy's minbpe.

    Jupyter Notebook

  2. llama2-rs llama2-rs Public

    Inference Llama 2 in many files of pure? Rust (karpathy's llama.c port).

    Rust

  3. zcat zcat Public

    Learn zig by implementing cat

    Zig

  4. gpt2-diy gpt2-diy Public

    From-scratch reproduction of GPT-2 following Andrej Karpathy's "Zero to Hero" series.

    Jupyter Notebook

  5. Moco Moco Public

    Moco contrastive unsupervised pretraining on Nih Chest X-Rays dataset.

    Python 2