Student & Aspiring AI Researcher | ICYS 2026 Gold Medalist | Building AI projects, researching independently and starting my career
AI-Atlas score
Strongest field
79 domain score
Global rank
Romania
| Machine Learning | 77 |
|---|---|
| Computer Vision | 79 |
| Data Science | 63 |
| NLP & LLMs | 73 |
| AI Agents | 28 |
| MLOps | 29 |
Each node is one domain, scored only from evidence in these public repositories. Select one to see why.
Deep, repeated, well-engineered work in this domain.
Ordered by how much each project adds to the Atlas score, not by its own score. A project with evidence in several domains can rank above one with a slightly higher project score. Choose a domain to focus on the projects with meaningful evidence there.
21% of Atlas Score
The repository provides a from-scratch PyTorch reimplementation of the ResNet family (ResNet-12, ResNet-18, ResNet-34, ResNet-50) including custom activation functions, batch normalization, training loops, and evaluation on CIFAR-100, accompanied by explanatory theory notebooks.
12% of Atlas Score
This repository provides a from-scratch PyTorch implementation of the LeNet-5 architecture, including a custom radial basis function output layer and loss function. It includes training and evaluation code, unit tests, and detailed educational documentation.
10% of Atlas Score
The repository contains a series of Jupyter notebooks and Python scripts that re-implement neural network concepts from scratch, following Andrej Karpathy's 'Neural Networks: Zero to Hero' course, covering bigram models, MLPs, RNNs, manual backpropagation, WaveNet-style architectures, character-level GPT, tokenization, and a GPT-2 reproduction.
9% of Atlas Score
ASTRO is a research project investigating modifications to the Muon optimizer for training language models. It implements the ASTRO optimizer and provides a benchmark harness to evaluate its performance against baselines.
9% of Atlas Score
This repository contains a collection of Jupyter notebooks that provide end-to-end solutions to machine learning competitions (MLCompete directory) and educational notebooks on topics such as CNNs and exploratory data analysis (Main directory). The competition solutions focus on feature engineering, model training, and generating submission files, while the educational notebooks offer step-by-step explanations of key concepts.
6% of Atlas Score
This repository is a collection of 38 runnable Jupyter notebooks that follow the Hugging Face LLM and Reasoning courses, with the author's own experiments and notes. It includes from-scratch implementations of tokenizers and GRPO, as well as an end-to-end Romanian question-answering project.
