47 repos across 2 sub-areas
Educational implementations and tutorials for building large language models and deep learning systems from scratch, primarily in PyTorch. Repositories here focus on demystifying transformer architectures, neural network fundamentals, and modern AI techniques through hands-on code examples and detailed explanations. While the cluster includes some specialized computer vision projects (NeRF, Gaussian splatting, SLAM), the dominant theme centers on learning-focused LLM and deep learning resources for practitioners seeking to understand these systems at a fundamental level.
Multimodal AI and Vision-Language Models
27 repos
Training, fine-tuning, and evaluation frameworks for large-scale vision-language models, video understanding systems, and multimodal AI. The cluster includes foundational work on world models for embodied AI, neural rendering techniques (NeRF), video representation learning, language model alignment via DPO, and specialized fine-tuning pipelines for models like Qwen2-VL. Repositories here focus on GPU-accelerated training infrastructure, model adaptation for downstream tasks, and the intersection of vision, language, and temporal reasoning.
Deep Learning & Computer Vision from Scratch
20 repos
Educational implementations of deep learning architectures and computer vision techniques, primarily in PyTorch. These repositories focus on learning-oriented implementations of foundational models—including attention mechanisms, pose estimation, shape reconstruction, and behavioral learning—with an emphasis on understanding how these systems work rather than production optimization. The cluster emphasizes hands-on, tutorial-style code that breaks down complex concepts like transformers, 3D shape representations, and multi-modal learning into approachable implementations.