Custom ML Model and Component Designer.
AI-Atlas score
| Machine Learning | 73 |
|---|---|
| Computer Vision | 65 |
| Data Science | 42 |
| NLP & LLMs | 69 |
| AI Agents | 33 |
| MLOps | 25 |
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.
23% of Atlas Score
The repository provides a fully fine-tuned 128-layer emotion classification model with CLI, API, and deployment tools. It includes training, evaluation, and inference optimization scripts, achieving 86.33% test accuracy on the DAIR.AI Emotion dataset.
22% of Atlas Score
MicroSR-X2 is a 2× super-resolution model with a custom quantized architecture (INT2 weights) distributed as a 495-byte .msr file. The repository provides inference code (CLI and GUI), optional training tools, benchmarks, and detailed documentation.
15% of Atlas Score
The repository releases a 14-million-parameter attention-free VPSM v5 language model for generating short NPC dialogue in Ultima Online. It includes the model checkpoint, tokenizer, runtime inference code, and two graphical user interfaces for interacting with the model.
14% of Atlas Score
Teach Me is an educational side-scrolling game where players train an AI via demonstration learning, featuring a custom PyTorch model, training loop, and game server that allows users to teach an AI to navigate procedurally generated levels.
13% of Atlas Score
UO-Mind NPC is a self-contained language model application designed for generating short, in-character replies for Ultima Online-style NPCs, featuring a custom 40M-parameter model with identity and profession locks, a language replacement filter, a desktop GUI, and a lightweight HTTP API.
13% of Atlas Score
The Float repository introduces a novel convolutional block (FloatingConvBlockV1) with Relative Energy Norm (REN) and a gating mechanism for stable training without running statistics, and provides a training script to evaluate it against a ResNet-18 baseline on CIFAR-10.
