https://diego-bonilla-portfolio.comProjects, research, articles and videos.https://diego-bonilla-portfolio.com/projects/biasloop/https://diego-bonilla-portfolio.com/projects/biasloop/Diffusion self-loop diagnostics: tracking how repeated model passes change image embeddings across strengths and visual content.https://diego-bonilla-portfolio.com/projects/continual-clip/https://diego-bonilla-portfolio.com/projects/continual-clip/A CLIP-like system that adds visual concepts directly in latent space from a handful of images without updating model weights.https://diego-bonilla-portfolio.com/projects/drawing-machine/https://diego-bonilla-portfolio.com/projects/drawing-machine/A DIY pen plotter with Arduino firmware and Python tools for manual control, image conversion and G-code output.https://diego-bonilla-portfolio.com/projects/drift-model-simulator/https://diego-bonilla-portfolio.com/projects/drift-model-simulator/An interactive 2D physics simulation of generative modeling via drifting, with positive, negative and generated particles moving in real time.https://diego-bonilla-portfolio.com/projects/fractal-tokenizer/https://diego-bonilla-portfolio.com/projects/fractal-tokenizer/Adaptive image tokenization using pattern mining and quadtree partitioning to create variable-size visual patches.https://diego-bonilla-portfolio.com/videos/how-to-make-your-llm-reason/https://diego-bonilla-portfolio.com/videos/how-to-make-your-llm-reason/A video about techniques for reasoning with language models.https://diego-bonilla-portfolio.com/research/orthogonality-explainability/https://diego-bonilla-portfolio.com/research/orthogonality-explainability/Tests whether orthogonal convolutional channels represent more independent visual concepts, with scoring and labels for attribution.https://diego-bonilla-portfolio.com/videos/ppo-and-grpo-in-llms/https://diego-bonilla-portfolio.com/videos/ppo-and-grpo-in-llms/A video explaining PPO and GRPO for language models.https://diego-bonilla-portfolio.com/research/region-coarsening-image-embeddings/https://diego-bonilla-portfolio.com/research/region-coarsening-image-embeddings/Progressively merge connected image regions and measure how Vision Transformer embeddings drift as pixel detail is removed.https://diego-bonilla-portfolio.com/videos/real-time-rgbd-slam/https://diego-bonilla-portfolio.com/videos/real-time-rgbd-slam/A short demonstration of mapping and camera tracking with a RealSense D415, OpenCV and Open3D.https://diego-bonilla-portfolio.com/projects/scenegraph-llm/https://diego-bonilla-portfolio.com/projects/scenegraph-llm/Zero-shot scene graph generation from images with LLM-supervised synthetic data and a vision-language relation model.https://diego-bonilla-portfolio.com/projects/slam-d415/https://diego-bonilla-portfolio.com/projects/slam-d415/An ongoing real-time mapping and camera-tracking project using an Intel RealSense D415 depth camera.https://diego-bonilla-portfolio.com/articles/maths-behind-sora/https://diego-bonilla-portfolio.com/articles/maths-behind-sora/An analysis of the architecture and techniques behind video generation.https://diego-bonilla-portfolio.com/articles/top-deep-learning-papers-2022/https://diego-bonilla-portfolio.com/articles/top-deep-learning-papers-2022/A review of deep learning papers from 2022.https://diego-bonilla-portfolio.com/articles/guide-to-transformer-architectures/https://diego-bonilla-portfolio.com/articles/guide-to-transformer-architectures/An explanation of transformer architectures and their components.https://diego-bonilla-portfolio.com/articles/give-me-120m-and-ill-change-ai/https://diego-bonilla-portfolio.com/articles/give-me-120m-and-ill-change-ai/An article about the cost of training a new language model.Tue, 07 Jul 2026 00:00:00 GMThttps://diego-bonilla-portfolio.com/library/ai-revolution/https://diego-bonilla-portfolio.com/library/ai-revolution/A coauthored textbook on deep learning theory and lab exercises.Tue, 28 Apr 2026 00:00:00 GMThttps://diego-bonilla-portfolio.com/articles/why-use-mse/https://diego-bonilla-portfolio.com/articles/why-use-mse/An explanation of mean squared error, from the practical intuition to its gradient and curvature.Mon, 10 Mar 2025 00:00:00 GMThttps://diego-bonilla-portfolio.com/articles/how-to-make-your-llm-reason/https://diego-bonilla-portfolio.com/articles/how-to-make-your-llm-reason/A written companion to the video series on reasoning, reinforcement learning and language models.Mon, 17 Feb 2025 00:00:00 GMThttps://diego-bonilla-portfolio.com/articles/tokenization-in-nlp/https://diego-bonilla-portfolio.com/articles/tokenization-in-nlp/How text is split into tokens and converted to model inputs.Mon, 11 Nov 2024 00:00:00 GMT