AIOSDirectory

JeffyAI

JeffyAI — free private AI assistant that runs on your own hardware. Your conversations never leave your machine.

JeffyAI is a privacy-first personal AI assistant built by Sagentica on the QAICX platform. Unlike cloud chatbots, it is designed to run locally on the user's own hardware so that prompts, files, and conversation history never leave the device. The product targets individuals and small teams who want a capable everyday assistant without sending data to a third-party server. It positions itself against hosted assistants by emphasizing offline operation, no per-message billing, and a free tier aimed at making private AI accessible to mainstream users.

QAIOS

QAIOS — AI operating system platform. Local-first, private by design, extensible with plugin bundles. Power users and SMBs.

QAIOS is an AI operating system platform from Sagentica that bundles a local large language model, a permission system, and a plugin architecture into a single private environment. It is built around a local-first philosophy: inference runs on the user's machine through models such as locally hosted Qwen, and capabilities are extended through modular plugin bundles rather than cloud services. The platform is aimed at power users and small and medium businesses that need an extensible AI workspace they can self-host, audit, and control without surrendering their data to an external provider.

Apple Intelligence

Apple's on-device AI system for iPhone, iPad, and Mac — Private Compute Cloud for privacy-preserving AI features.

Apple Intelligence is Apple's personal intelligence system built into iOS, iPadOS, and macOS on supported devices. It runs generative models directly on-device for tasks such as writing assistance, summarization, image generation, and a more capable Siri, falling back to Apple's Private Cloud Compute only when a request needs larger models. Private Cloud Compute extends on-device privacy guarantees to Apple silicon servers, where Apple states that user data is never stored and is cryptographically inaccessible even to Apple. The system illustrates the consumer push toward hybrid on-device and private-server AI integrated at the operating system level.

Google Gemini Nano

Google's smallest Gemini model for on-device inference — runs locally on Android devices for offline AI features.

Gemini Nano is the most compact member of Google DeepMind's Gemini model family, engineered to run directly on mobile and edge devices. It powers on-device features in Android through the AICore system service, enabling capabilities such as summarization, smart reply, and proofreading without sending data to the cloud. First shipped on Pixel devices and later extended to other Android hardware, Gemini Nano lets developers access generative AI offline with lower latency and improved privacy. It exemplifies the trend of embedding small, efficient foundation models into the consumer operating system itself.

Microsoft Copilot+

Microsoft's Copilot+ PC initiative — NPU-accelerated AI features built into Windows for next-gen hardware.

Galaxy AI is Samsung's suite of artificial intelligence features integrated into its Galaxy smartphones, tablets, and wearables. It includes Live Translate for calls, Interpreter mode, Chat Assist for tone adjustment, Circle to Search, and generative photo editing tools, combining on-device processing with cloud models. Samsung offers an on-device processing toggle so privacy-conscious users can restrict features to local computation. Built on top of Android in partnership with Google, Galaxy AI is one of the most widely distributed consumer AI experiences, bringing assistant and generative capabilities to hundreds of millions of mobile devices worldwide.

Samsung Galaxy AI

Samsung's Galaxy AI features — on-device and cloud AI for Galaxy smartphones and tablets.

MLflow is an open-source MLOps platform created by Databricks for managing the end-to-end machine learning lifecycle. It provides four core components: experiment tracking for logging parameters and metrics, a packaging format for reproducible runs, a model registry for versioning and stage transitions, and deployment tooling for serving models across environments. More recent releases add tracing and evaluation features aimed at generative AI and LLM applications. Framework-agnostic by design, MLflow integrates with scikit-learn, PyTorch, TensorFlow, and Spark, and is widely adopted by data science teams that need an auditable record of model development.

MLflow

Open-source MLOps platform — experiment tracking, model registry, deployment, and evaluation in one tool.

Ray is an open-source unified framework, originally developed at UC Berkeley's RISELab and now stewarded by Anyscale, for scaling Python and AI applications from a single laptop to a large cluster without rewriting code. Its libraries cover distributed training, hyperparameter tuning, reinforcement learning, batch inference, and model serving through components such as Ray Train, Ray Tune, RLlib, and Ray Serve. Many large-scale machine learning systems, including parts of well-known LLM training pipelines, rely on Ray to coordinate compute, making it a foundational layer for teams building distributed AI workloads.

Ray

Open-source framework for distributed AI and Python workloads — scale from laptop to cluster with one framework.

Clear Linux OS is a rolling-release Linux distribution developed by Intel and tuned aggressively for performance on Intel hardware. It uses aggressive compiler optimizations, function multi-versioning, and a stateless design where system configuration is cleanly separated from the OS image. The distribution ships optimized math and deep learning libraries, making it a common choice for benchmarking and running AI and high-performance compute workloads where squeezing maximum throughput from the processor matters. For teams deploying inference or training close to the hardware, Clear Linux serves as a minimal, performance-oriented base operating system.

Clear Linux

Intel's performance-optimized Linux distribution — tuned for AI workloads with hardware-specific optimizations.

Talos Linux is a minimal, immutable, and secure Linux operating system built by Sidero Labs specifically to run Kubernetes. It has no SSH, shell, or package manager; the entire system is managed through a declarative API and gRPC, which dramatically reduces its attack surface. The OS runs entirely in memory from a read-only image and is configured with a single machine configuration file, making clusters reproducible and easy to upgrade. For teams running AI and machine learning workloads on Kubernetes at the edge or in data centers, Talos provides a hardened, API-driven foundation with minimal operational overhead.

k3s + KubeAI

KubeAI on k3s — lightweight Kubernetes-native AI inference platform for self-hosted model serving.

Ubuntu remains the most widely used Linux distribution for artificial intelligence and machine learning development, and Canonical packages a dedicated AI stack and reference toolchain around it. The offering bundles optimized drivers, container tooling, MLOps components such as Charmed Kubeflow, and certified hardware support for NVIDIA, Intel, and other accelerators across edge, cloud, and workstation deployments. Canonical also provides long-term security maintenance through Ubuntu Pro for production AI infrastructure. For organizations standardizing their machine learning environment, Ubuntu for AI offers a supported, enterprise-grade operating system foundation with a well-known ecosystem of frameworks and drivers.

Talos Linux

Immutable Linux OS designed for Kubernetes — minimal attack surface, ideal for running k8s AI workloads.

Modal is a serverless cloud platform that lets developers run Python code, including GPU-accelerated AI and machine learning workloads, without managing infrastructure. Developers define container images, hardware requirements, and scaling rules directly in Python, and Modal provisions GPUs on demand and bills per second of usage. It is widely used for batch inference, model fine-tuning, large-scale data processing, and deploying inference endpoints, with fast container cold starts as a key selling point. Modal targets AI engineers who want cloud GPU compute that feels like calling a local function, removing the friction of Kubernetes and cloud provisioning.

TrueNAS SCALE + AI

TrueNAS SCALE with AI workload support — Kubernetes-native storage OS for self-hosted AI and ML pipelines.

Replicate is a cloud platform that lets developers run and deploy open-source and custom AI models through a simple API without managing GPU infrastructure. Thousands of community-published models for image generation, language, audio, and video can be called with a single HTTP request, and users package their own models using the open-source Cog tool. Replicate bills by the second of compute used and scales automatically with demand. It is popular with product teams and indie developers who want to integrate state-of-the-art open models into applications quickly, treating model inference as a managed API rather than self-hosted infrastructure.

Ubuntu for AI

Canonical's Ubuntu AI stack — curated OS and toolchain for AI workloads on edge, cloud, and workstation hardware.

NVIDIA AI Workbench is a free, locally installable toolkit that helps data scientists and developers create, collaborate on, and migrate GPU-accelerated AI projects across workstations, data centers, and the cloud. It packages work into containerized, version-controlled project environments so models and dependencies move cleanly between a local RTX workstation and remote GPU servers. The tool integrates with Git and container registries and ships example projects for fine-tuning and retrieval-augmented generation. It targets practitioners who want reproducible, portable AI development environments without manually configuring CUDA, drivers, and container stacks on every machine.

Modal

Serverless cloud platform for AI and ML workloads — run Python functions on GPUs in seconds with zero infra setup.

Replicate

Run open-source AI models via API — simple, scalable model hosting with per-second billing.

NVIDIA AI Workbench

NVIDIA's AI development environment for data scientists — standardized GPU workspace for training and fine-tuning.