My curated hub for learning, building, and exploring ML—locally, in-browser, and beyond.
For those starting out or exploring advanced tools, this list covers frameworks, open-source models, browser-based AI tools, and courses that are completely free to use.
A JavaScript library for training and running ML models in the browser or Node.js. Great for tasks like image classification and NLP using pre-trained models.
ONNX.js
Run pre-trained ONNX models in the browser using WebAssembly (WASM) for performance.
Transformers.js
Run transformer models like BERT or GPT-2 directly in the browser with WebAssembly. Useful for privacy-sensitive NLP apps.
ml5.js
Beginner-friendly library built on TensorFlow.js. Offers APIs for image classification, pose detection, and more.
Web AI
Lightweight library for running deep learning models in-browser. 👉Demo Site
Local Machine Learning Platforms
Ollama
Run large language models locally with privacy and performance in mind. Supports models like GPT-J, LLaMA, and Mistral.
Page Assist
Run local models directly in your browser (Chromium or Firefox) as an extension.
RunPod
GPU-powered pods for running ML workloads in the cloud or locally. Supports NVIDIA A100/H100 GPUs.
Open WebUI
Offline UI solution for running models like Ollama locally. Supports OpenAI-compatible APIs.
General Open-Source ML Tools
MLflow
Track experiments, manage models, and deploy across platforms.
Gradio
Create easy-to-share web apps for ML models with multiple input types (text, image, audio).
Streamlit
Convert Python scripts into interactive web apps—no front-end skills needed.
Hugging Face Transformers
Popular library for pre-trained models in NLP, vision, and audio.
Specialised Tools & Open-Source Models
Specialised Tools
ONNX Runtime Web: High-performance inference engine using WebAssembly and WebGL for GPU acceleration in the browser.
Vertex AI (Google Cloud): Managed platform to build and scale ML applications with BigQuery and LangChain integration.
Google AI Studio: Web-based workspace for experimenting with Google's Gemini AI models.
Open-Source Models
Google Gemma 2: Open LLM with 9B and 27B parameter variants, optimised for different hardware.
BLOOM: Multilingual model supporting 46 human and 13 programming languages.
GPT-NeoX: 20B parameter autoregressive LLM trained on The Pile.
DeepSeek R1: Reasoning-focused LLM with a 128K token context window and multilingual capabilities.
LLaMA : Meta's high-performance open-source LLM for research and deployment.