This daily observation tracks 24 AI deployments across manufacturing, healthcare, agriculture, transport, finance, governance, and terminals on Nov 15, 2026, highlighting embodied AI quality inspectors, flexible production lines, agricultural robots, medical agents, financial multi-agent systems, city-scale intelligent bodies, AI glasses, and token economy infrastructure.
This November 2026 daily observation analyzes 25 AI+industry cases across manufacturing, healthcare, agriculture, transport, finance, governance, and rural sectors, revealing a shift from pilot projects to agent cluster collaboration and data flywheel growth, with detailed metrics on efficiency gains, cost reductions, and deployment scale.
A daily observation of 24 AI+industry cases from Nov 1–11, 2026 reveals agents moving beyond Q&A to autonomous execution in finance, healthcare, and urban governance, while industrial clusters, vehicle-road-cloud, rural deployment, and edge AI hardware accelerate real-world adoption.
This report analyzes 23 AI application cases across manufacturing, healthcare, agriculture, transportation, finance, governance, terminals, and rural sectors from Nov 1-10, 2026, highlighting the shift from pilot projects to autonomous agent clusters, data flywheels, and commercialized vehicle-road-cloud integration, with AI dividends reaching county-level villages and edge hardware launching a new spatial computing cycle.
This report tracks AI industry applications from Nov 1-9, 2026, highlighting the shift from pilot projects to multi-agent collaboration and data flywheels across manufacturing, healthcare, agriculture, transport, finance, governance, and edge hardware, with concrete metrics and case studies.
This daily observation covers AI applications across manufacturing, healthcare, agriculture, rural governance, transportation, finance, social governance, and consumer terminals on November 5, 2026, highlighting five key trends: multi-agent collaboration, rural AI penetration, vehicle-road-cloud commercialization, edge AI hardware, and capital-driven industrial investment, with concrete metrics from real-world deployments.
This daily observation tracks 21 benchmark cases across 10 sectors where AI agents have moved from pilot to autonomous execution, delivering measured gains in manufacturing, healthcare, transportation, finance, and governance, while highlighting trends like county-level adoption, compliance frameworks, and edge-device hardware cycles.
AI agentsautonomous executioncompliance frameworks
An engineering teardown of the Arrowspace S1 AI screwdriver reveals how a 2 MP camera, polarized lighting, and a 0.5 TOPS NPU identify screw types in milliseconds, while a 1 kHz strain-gauge sensor and PID loop achieve ±5% torque accuracy — all within a 0.5 W compute budget powered by a single 18650 cell.
This daily AI industry report analyzes 19 real-world deployments across manufacturing, healthcare, agriculture, transportation, finance, governance, terminals, investment promotion, and rural management, showing AI agents transitioning from pilots to scalable production with quantified metrics like 95% temperature accuracy in steelmaking, 30% faster diagnosis in county hospitals, and 15% traffic delay reduction in Beijing.
This article explains embedded systems as specialized computers embedded in devices like washing machines, cars, and smartwatches, covering their hardware components, software architectures, real-time requirements, development tools, and emerging edge AI applications.
This daily observation tracks 18 AI deployments across 10 industries including manufacturing, healthcare, agriculture, and transportation, revealing key trends: AI agents are maturing from standalone tools into integrated systems that execute tasks, benefits are reaching county and rural areas, compliance frameworks are emerging, and edge AI hardware is triggering a new device replacement cycle.
A daily observation covering 19 latest AI deployment cases across manufacturing, healthcare, agriculture, rural governance, transportation, finance, social governance, terminals, and investment promotion, showing AI shifting from advisory to operational roles with industrial agents entering factory cores, government agents handling 2M+ citizen requests, and on-device 30B models enabling offline complex reasoning.
This daily observation tracks 14 AI deployments across manufacturing, healthcare, agriculture, transport, finance, governance, rural development, terminals, and investment promotion, revealing a shift from conversational AI to autonomous agents, with data flywheels and domain knowledge as key enablers.
At the 2026 Yunqi Conference, Banma Smart unveiled AutoOmni 2.0-23B-A3B, a sparse MoE edge model that achieves near-cloud performance on vehicle-grade chips, addressing memory, latency, and concurrency constraints while enabling privacy-preserving, low-cost AI with certified safety frameworks for mass production.
OlliteRT is an open-source Android app that runs large language models locally via Google's LiteRT runtime, exposing an OpenAI-compatible API over LAN for lightweight always-on tasks like RSS summarization and smart home automation, with monitoring, logging, and security features.
This article argues that AI models are transient compute resources while user memory is a persistent asset, analyzing MemVerge's MemoryBox as a local-first memory layer that separates memory from models, enables multi-model scheduling, and addresses data sovereignty amid the shift to edge AI.
OlliteRT is an open-source Android app that transforms old phones into local LLM servers exposing OpenAI-compatible APIs, supporting models like Gemma 4 E2B and Qwen 2.5 1.5B with configurable hardware acceleration, monitoring, and Home Assistant integration, all running offline on 5-10W power.
The article breaks down the three‑layer foundation—compute unlocking with Arm SME2, model slimming via Tencent's ultra‑low‑bit quantization, and deployment using ExecuTorch and Windows on Arm—that enables large AI agents to operate efficiently on mobile, PC, and other edge devices.
A joint test by Diffusion Intelligence and Acrab shows that diffusion language models can speed up edge‑side agents by about five times compared to traditional autoregressive models, thanks to parallel token generation and global‑content framing, promising faster, more flexible AI experiences on devices.
AI accelerationagent inferenceautoregressive models
The article examines how large‑model agents can finally run on phones, PCs, cars and IoT devices by tackling three engineering layers—unlocking compute with Arm SME2, slimming models through ultra‑low‑bit quantization, and deploying with ExecuTorch and Windows on Arm—presented at the free Arm Create 2026 events.
The Arduino VENTUNO Q, priced at $299 and shipped with Ubuntu, combines a Qualcomm Dragonwing IQ‑8275 NPU delivering up to 40 TOPS for on‑device large‑language‑model inference with an STM32H5 MCU handling sub‑millisecond real‑time control, offering integrated support for offline voice assistants, multi‑camera vision, ROS 2 robotics, and industrial CAN‑FD interfaces, positioning it as a ready‑to‑use edge AI platform distinct from Raspberry Pi or generic Ubuntu PCs.
Needle2 tackles the twin constraints of cloud‑dependency and heavyweight local models by offering a 45 M‑parameter, 14 MB single‑file model that runs in a constant ~28 MB memory footprint, specializing in tool calling, device control, and JSON extraction on ultra‑light hardware such as Raspberry Pi, VR headsets, and ESP32‑S3, while providing detailed architecture, benchmark, deployment, and fine‑tuning guidance.
This article reviews the ten fastest‑growing AI open‑source projects on GitHub over the past week, detailing each project's core capabilities, technical architecture, and ecosystem impact while highlighting three emerging trends: AI agents becoming production tools, the rise of edge‑centric lightweight deployment, and accelerated open‑source contributions from major tech firms.
The 2026 intelligent computing center report analyzes the shift from training‑centric GPU stacks to heterogeneous AI super‑factories dominated by inference, outlines new cooling and optical technologies, and explains how national policies drive a cloud‑edge hierarchy focused on utilization, energy efficiency, and domestic ecosystem development.
The paper presents Jetson-PI, an open‑source VLA real‑time control framework for low‑power edge devices that tackles inference latency and perception‑action misalignment through foresight‑aligned asynchronous correction, confidence‑based scheduling, and edge‑engine optimizations, raising Jetson Orin control frequency from 0.7 Hz to 6.06 Hz and improving task success rates on LIBERO benchmarks and a real‑world clothing‑folding robot.
Apple’s July 2026 public beta of Siri replaces its legacy intent‑based pipeline with a large‑language‑model‑driven agent architecture, introducing multimodal perception, persistent memory, and a three‑tier edge‑cloud inference system that reshapes mobile assistants while emphasizing privacy through on‑device processing and differential‑privacy techniques.
PrismML’s Bonsai compresses the 27B Qwen3.6 model to a 3.9 GB 1‑bit version that runs on iPhone 17 Pro and in browsers via WebGPU, achieving competitive benchmark scores, high "intelligent density", and enabling zero‑cost, privacy‑preserving agent applications on edge devices.
Om AI announced a three‑model VLX suite—VLX‑Flow, VLX‑Seek and VLX‑Go—designed to keep video streams continuously feeding a device‑side brain, using incremental visual memory and linear attention to meet the low‑latency, resource‑constrained demands of real‑world cameras, drones and robots.
The article examines the 2025‑2026 surge of sub‑10B parameter models, outlines edge‑side AI agents, details multimodal interaction layers, and explains how Computer Use enables AI to operate computers, offering a roadmap for integrating these technologies across mobile, desktop, and IoT scenarios.
The article examines why emerging wearables, smart glasses, and in‑car systems need a "first‑layer" on‑device AI that preprocesses multimodal inputs, outlines the missing input, application, and system capabilities required for edge models, and discusses how subsequent edge and cloud stages should share the workload.
Google DeepMind’s Gemma 4 12B model, released under Apache 2.0, runs fully offline on a 16 GB laptop, uses a novel no‑encoder unified architecture, delivers 80 token/s with only 9 GB VRAM, and matches the quality of the 26 B predecessor while powering advanced agentic and multimodal demos.
The article examines Nvidia’s RTX 5090 Ti and Project DIGITS announcements, showing how desktop GPUs and AI engines now enable trillion‑parameter models on laptops, and analyzes the resulting shift from cloud‑centric AI to edge computing, including cost, latency, data‑sovereignty benefits and the challenges enterprises face.
This article reviews the ten fastest‑growing open‑source AI projects on GitHub over the past week, detailing each project's core capabilities, architecture, and impact while highlighting three emerging trends: AI agents becoming production tools, the rise of edge and lightweight deployments, and accelerated open‑source contributions from major tech firms.
Alibaba International hosted an Academic Bar at Zhejiang University where experts presented multilingual e‑LLM models, next‑generation recommendation systems, AI‑driven optical design, safety frameworks for agents, a benchmark for e‑commerce chatbots, virtual user research, an Agent Harness architecture, and edge‑side large‑model techniques, illustrating the practical convergence of cutting‑edge AI research and industry.
Samsung is rapidly building its own foundation model ecosystem, introducing the memory‑based Meki architecture to leverage ROM for edge AI, the multi‑domain M2RL reinforcement‑learning paradigm, and the LiveClawBench 3‑dimensional benchmark to evaluate physical‑world AI, signaling a strategic shift from cloud‑centric to physical AI deployment.
The 1‑billion‑parameter MiniCPM5-1B model tops the AA leaderboard with a 17.9 score, outperforms 2‑billion‑parameter rivals, uses an AI‑generated training framework that cuts cost by 10%, and runs on virtually any device thanks to aggressive quantisation and open‑source tooling.
The article analyzes MiniCPM5‑1B, a 1‑billion‑parameter edge‑friendly language model whose training framework, ForgeTrain, was generated entirely by AI, achieving Megatron‑level quality with 10% faster speed and enabling low‑cost, low‑latency deployment on devices ranging from laptops to smartphones.
EUPE introduces a three‑stage “scale‑then‑shrink” distillation pipeline that first trains a large proxy model to absorb heterogeneous expert knowledge and then compresses it into an 86M encoder, achieving state‑of‑the‑art performance on image classification, dense prediction and vision‑language tasks on an iPhone with only 62 ms latency.
MiniCPM‑V 4.6, a 1.3 B‑parameter multimodal LLM, outperforms larger rivals such as Qwen3.5‑0.8B and Gemma 4 on both accuracy and speed, thanks to early ViT token compression and 4×/16× visual token reduction, delivering sub‑100 ms latency and over 2.6 k token/s throughput on a single RTX 4090 while also running offline on mobile devices.
Running a 400‑billion‑parameter model locally on the iPhone demonstrates a leap in model compression and edge AI, turning the device into a cognitive agent that handles tasks without apps, while Apple’s upcoming iOS 27 visual‑intelligence features and hardware upgrades cement its role as the core AI ‘digital passport’.
The article examines the limitations of traditional transparent‑forwarding satellite links, proposes a dual‑engine "communication + AI" architecture for satellite‑borne AI base stations, explores resource‑pooling, space‑app‑store micro‑services, and real‑world use cases in wildfire detection, maritime navigation and renewable‑energy grid management, and outlines the path toward 6G‑enabled space‑ground computing networks.
This roundup analyzes ten rapidly rising open‑source AI projects—covering self‑evolving agents, multimodal models, edge deployment, and quantum AI—highlighting their technical innovations, benchmark results, and emerging industry trends that are reshaping AI development and deployment.
A Chinese research team’s 2024 "density law"—which predicts that the parameters needed for a given LLM performance halve every 3.5 months—has been independently validated by Meta’s scaling ladder, METR’s time‑horizon report, and subsequent analyses, revealing a unified exponential growth curve that reshapes expectations for inference cost, edge AI feasibility, and optimal model‑development strategies.
Edge AI runs AI models directly in browsers or devices, offering zero latency, zero API cost, and full privacy, and the article explains the three technical breakthroughs that make it possible, compares WebLLM, Transformers.js and Ollama, and provides a hybrid architecture with concrete engineering challenges and solutions that can cut total AI costs by 40‑55% for typical front‑end applications.
Google AI Edge introduces two open‑source projects—Gallery and LiteRT‑LM—that let Android developers run large language models locally without network connectivity, offering offline inference, privacy protection, GPU/NPU acceleration, and streaming output for real‑time AI experiences.
Google’s open‑source AI Edge Gallery lets Android and iOS devices run large language models such as Gemma 4 entirely offline, eliminating network latency and privacy concerns; the app showcases six modular AI features, offers a simple install path, and signals Google’s push toward a standardized edge‑AI ecosystem.
Google’s newly released Gemma 4 series delivers a range of open‑source LLMs—from 2.3 B to 31 B parameters—optimized for edge devices through per‑layer embeddings, mixed‑expert MoE, hybrid attention, and extensive hardware support, achieving top‑tier benchmark scores while running efficiently on phones and IoT.
The March 2026 AI landscape features a 2.0 era of open‑source large models led by DeepSeek‑R1, a breakout year for AI Agents with hierarchical planning and robust tool calls, and a cost‑driven showdown among GPT‑5.4, Claude Opus 4.6 and Gemini 3.1 Pro, reshaping capabilities, pricing, and deployment strategies across cloud and edge.
RuView, an open‑source edge‑AI project built on a CMU Wi‑Fi DensePose paper, claims to achieve wall‑penetrating human pose estimation, vital sign monitoring, and ultra‑fast presence detection using only standard Wi‑Fi signals and low‑cost ESP32 hardware, while sparking intense community debate over its claimed capabilities and reproducibility.
By introducing a dynamic mixture‑of‑experts routing scheme and an end‑to‑end architecture that eliminates NMS and DFL, YOLO‑Master and YOLO26 dramatically cut compute waste and latency on edge devices, achieving up to 43% faster CPU inference while keeping model accuracy, with all code openly released.
The article traces digital twin evolution from early 2018‑2021 mirror simulations through real‑time perception to the emerging self‑healing closed‑loop in 2025‑ onward, detailing the four‑layer architecture, key 2026 technologies, real‑world deployments, and remaining challenges for manufacturers.
Alibaba released four Qwen‑3.5 models (0.8B‑9B) that use a Gated DeltaNet hybrid‑attention architecture and native multimodal training to achieve 262k‑token contexts, outperform larger rivals on visual, reasoning, and math benchmarks, and run video analysis on phones and laptops, though they still demand significant VRAM.
The article explains how the Taalas HC1 edge‑AI chip, with 17,000 tokens/s inference speed, 90 % lower power and 1/20 the cost of Nvidia H200 GPUs, proves that dedicated, non‑general‑purpose silicon can overcome latency, privacy and expense barriers, making on‑device large‑model deployment essential in 2026 and offering a strategic roadmap for Chinese chip makers.
The article outlines ten actionable digital trends for 2026—from AI agents and large‑model inference cost strategies to platform engineering maturity, FinOps‑driven multi‑cloud governance, zero‑trust deepening, edge AI deployment, data contracts, and value‑centric tech organization—explaining why each matters and how CTOs can quickly build competence.
The article showcases three low‑cost edge deployments of OpenClaw—a $25 Moto phone using Termux, a $5 ESP32‑S3 board running a pure‑C MimiClaw, and the BotDrop Android app that turns an old phone into an AI agent host—detailing setup steps, challenges, and security considerations.
AgentCPM‑Explore, a 4‑billion‑parameter open‑source agent model, breaks the conventional belief that larger models always perform better by achieving state‑of‑the‑art results on eight long‑duration benchmarks, surpassing many 8B and even some 30B models while enabling efficient edge deployment.
YOLO26, the latest Ultralytics release, introduces a unified model family with five sizes, removes distribution focal loss, offers end‑to‑end inference without NMS, adds progressive loss balancing and the MuSGD optimizer, and delivers up to 43% faster CPU performance, making it ideal for edge and real‑world vision applications.
The article analyzes how AI in 2026 will shift from conversational hype to actionable agents, featuring paradigm changes toward act‑oriented agents, a split between edge‑efficient and slow‑thinking models, deep multimodal fusion, and embodied intelligence that turns AI into a practical digital colleague.
The article analyzes the severe inference bottlenecks of large language models, CNNs, and recommendation systems and presents a suite of research‑driven accelerations—including token‑level pipeline parallelism (HPipe), KV‑cache clustering (ClusterAttn), quantization (QoKV), heterogeneous edge frameworks (DeepZoning, PICO), delay‑aware edge‑cloud scheduling (DECC), and operator choreography (RACE)—validated on real‑world industrial workloads.
Mistral AI’s latest open‑source release, Mistral 3, introduces three compact dense models and the powerful Mistral Large 3 MoE model, outperforming domestic rivals in benchmarks, offering strong multilingual and multimodal capabilities, and delivering the lowest cost‑performance ratio among open‑source LLMs.
LocalAI is an open‑source, self‑hosted alternative to OpenAI that lets you run large language, image and audio models on your own CPU or GPU, offering full data privacy, zero cloud costs, and offline capability while remaining compatible with the OpenAI API ecosystem.
This article analyzes the technical foundations, development frameworks, real‑time inference optimizations, typical industry deployments, and future research directions of multi‑agent systems in 2025, highlighting protocols like FIPA‑ACL and MCP, tools such as LangGraph and ADP3.0, and edge‑computing breakthroughs.
nndeploy is an open‑source, high‑performance on‑device AI deployment framework that abstracts the repetitive “last‑mile” workflow into a visual drag‑and‑drop DAG, offering multi‑platform inference, optimization, and ready‑to‑use model configs, enabling developers to go from prototype to production in minutes.
The 2026 China AIoT Industry Annual Conference gathered over 200 experts and 600,000 online viewers to unveil AIoT 2.0 trends, from mobile IoT breakthroughs and edge AI solutions to global connectivity, Web3 integration, and award‑winning innovations shaping the future of intelligent connectivity.
This comprehensive analysis examines the strategic importance, technical challenges, innovation pathways, and market landscape of domestic AI chips in China, highlighting key players, regional clusters, core applications such as intelligent computing, autonomous driving, and robotics, and projecting future industry bottlenecks and opportunities.
At the 2025 Vivo Developer Conference, Baidu Security earned the Best Security Technology Partner award for its edge‑focused large‑model security solution, which tackles multi‑layered threats on devices through comprehensive content protection, tailored edge defenses, advanced attack detection, and a rigorous evaluation framework.
This article explains the principles, key methods, and practical effects of model quantization, pruning, and knowledge distillation, comparing their advantages and disadvantages, and showing how combining these techniques enables compact, high‑performance AI models on resource‑constrained devices.
The article examines how the rapid growth of AI models and workloads is reshaping network design, highlighting the need for ultra‑high bandwidth, sub‑millisecond latency, reliability, scalable topologies like Fat‑Tree and Dragonfly, and robust security and QoS mechanisms across data‑center, cloud, and edge environments.
Google’s newly released Gemma 3 270M is a compact 270‑million‑parameter language model that combines a large token vocabulary, energy‑efficient INT4 quantization, strong instruction‑following, and production‑ready checkpoints, making it ideal for fine‑tuning, on‑device deployment, and a wide range of low‑latency AI tasks.
Google has released the open‑source Gemma 3 270M model—a compact, 270‑million‑parameter AI that runs on as little as 2 GB RAM, supports over 140 languages, handles images, and offers strong instruction‑following performance, making it ideal for edge devices and custom fine‑tuning.
The newly released “Terminal Agent Security 2025” report, unveiled at the World AI Conference, systematically categorizes AI agent risks, outlines detection and defense methods, and proposes three protection pathways—single‑agent safety, trustworthy multi‑agent interconnection, and AI‑terminal security—to guide the emerging ecosystem of intelligent edge devices.
This article explains why quantizing massive AI models is essential for deploying them on resource‑constrained devices, outlines core quantization concepts, techniques, and methods, compares their pros and cons, and presents practical application scenarios such as smartphones, autonomous driving, IoT, and edge computing.
The article details how edge intelligence is applied to the Vivo official app to improve product recommendation on the smart‑hardware floor by abstracting the problem, designing feature engineering pipelines, training TensorFlow models, converting them to TFLite, and deploying inference on mobile devices, while also covering monitoring and performance considerations.
This article explains how Gaode Map leverages lightweight edge TTS models, dual‑autoregressive large‑model data augmentation, and a configurable audio‑processing DAG to enable users to create highly realistic personalized voice packs from just three recorded sentences.
This article details how Vivo’s official app implements edge intelligence to dynamically rank and recommend hardware products on its homepage, covering problem abstraction, data collection, feature engineering, model design, TensorFlow‑Lite conversion, on‑device inference, and monitoring for a personalized user experience.
This paper presents HGO‑YOLO, a lightweight real‑time anomaly‑behavior detector that integrates HGNetv2 and GhostConv into YOLOv8, achieving 87.4% mAP with just 4.6 MB of parameters and 56 FPS on CPU, and validates its performance across multiple datasets and hardware platforms.
This article provides an in‑depth technical analysis of Cambricon’s AI chip portfolio—including terminal, cloud, and edge processors—detailing their micro‑architectures, key innovations such as chiplet technology and memory optimisation, roadmap plans, and real‑world applications in data centers, surveillance and autonomous driving.
This article provides a detailed technical analysis of Huawei's Ascend 310 AI processor, covering its core compute performance, power efficiency, DaVinci architecture, video/image processing capabilities, recent product roadmap, and the broader industry impact of Huawei's AI chip ecosystem.
The article analyzes the challenges of deploying large language models on cloud servers—such as latency, security, and constant connectivity—and explains how near‑memory computing architectures (PNM, PIM, CIM) can integrate storage and processing to enable efficient, high‑performance edge AI deployments, outlining the trade‑offs of each approach.
The article describes an edge‑based video‑cover recommendation system for DeWu that leverages the MNN SDK and a lightweight MobileNetV3 model, performing on‑device inference with quantization and parallel processing to automatically select high‑quality covers, achieving sub‑second latency and boosting click‑through rates by up to 18 %.
The article outlines ten major AI trends expected in 2025, including the rise of intelligent agents, fierce competition among multimodal large models, breakthroughs in text‑to‑video generation, long‑term memory, quantum computing, edge models, embodied intelligence, humanoid robots, AI self‑looping with synthetic data, and the enduring scaling laws of large models.
This article explores how compact multimodal models like OmniVision-968M enable efficient generative AI on edge devices, detailing their architectural advantages, benchmark superiority over larger models, and step‑by‑step instructions for local installation and visual inference using NexaSDK.
The article examines the rapid growth and increasing complexity of edge AI NPUs, discussing challenges in software and hardware acceleration, supply‑chain constraints, and the need for integrated engine solutions to sustain performance and power efficiency.
Google I/O highlighted the rise of on‑device AI, showing how new neural processors, Edge TPU, and tools like the Edge AI SDK and TensorFlow Lite enable developers to run large language models locally, reducing latency, cost, and privacy concerns while integrating with cloud resources.
This article presents OPPO's practical research on deploying multi‑modal pre‑training models across mobile devices and cloud, covering edge image‑text retrieval, text‑image generation and understanding optimizations, and lightweight diffusion model techniques, with detailed algorithmic improvements, performance results, and real‑world application cases.
The article analyses the 2024 AI landscape, highlighting a multimodal explosion, the limits of current AI applications, Sora as a concrete step toward AGI, the rise of AI‑native business models, edge‑AI hardware opportunities, the challenges of human‑level models, and the broader societal impacts of an AI‑driven data era.
This article details JD Retail's edge AI (on‑device intelligence) platform, covering its definition, performance and security challenges, three‑layer cloud‑edge‑device architecture, key components such as high‑performance inference engine, data pipeline, Python VM container, and real‑world applications in traffic distribution and image recognition.
This tutorial walks you through installing Ollama on a Raspberry Pi, exploring TinyLlama, Phi, and LLaVA models, and demonstrates how to run and interact with these LLMs locally, including hardware requirements and practical command examples.
This article examines the rapid rise of edge AI video analytics, explains how Vision Transformers (ViT) overcome the limitations of traditional CNNs, details a technical pre‑research and POC conducted by a Chinese AI firm, evaluates several open‑source large models, and concludes that the OFA model best meets current edge deployment needs.
MetaSight introduces a real‑time video‑stream subject recognition system that replaces the traditional capture‑upload‑search flow with continuous, on‑camera product identification, using a sub‑10 MB edge model, global IDs for frame continuity, edge‑cloud collaboration, and batch processing to cut interaction steps, lower server load, and pave the way for future AR/XR shopping experiences.
This article explains how Alipay’s homepage team collaborates with the edge‑intelligence team to use real‑time client‑side behavior data and algorithm platforms, transforming refresh strategies across time, space, and event dimensions, improving recommendation efficiency, reducing duplication, and delivering measurable performance gains.
This article reviews Alipay's mobile visual algorithm ecosystem, detailing its diverse application scenarios, technical challenges, architectural framework, lightweight design strategies, scalable modeling techniques, and future research directions for edge AI on billions of devices.
This article explains the evolution of Ant Group's xNN edge‑AI framework, detailing its four‑layer model‑optimization space, the lightweight modeling of version 1.0, and the transition to scalable modeling in version 2.0 to better exploit fragmented device compute resources.
This article explains the concept of edge intelligence, its advantages over cloud‑based AI, the technical challenges of deploying AI on mobile devices, Ant Group's development timeline, core technology stack, and future directions for edge‑cloud collaboration.
This article examines the rapid growth of edge AI, outlines its current development, identifies technical and deployment challenges, presents key innovations such as RepRetinaFace and DeepStream, and showcases a 5G‑enabled smart construction solution with concrete performance data and implementation details.
Pitaya, built by ByteDance’s Client AI and MLX teams, is a comprehensive end‑side AI engineering platform that provides a full workflow from model development and data preparation to deployment, monitoring, and federated learning, supporting large‑scale commercial scenarios across multiple apps.
The article demonstrates how to evaluate, compress, and convert deep‑learning models for edge devices using TensorFlow, JAX, and TVM—showing a faster iPhone‑based MNIST training benchmark, FLOPs measurement scripts, TFLite/ONNX/CoreML conversion, TVM compilation with auto‑tuning, and up to 50 % speed improvements on mobile NPU hardware.
This article examines the computational capabilities of modern mobile devices for machine learning, compares training times on a MacBook and iPhone, explains model evaluation metrics like FLOPs, and provides step‑by‑step guides for converting and optimizing models using TensorFlow, PyTorch, ONNX, JAX, and TVM for edge deployment.
This article details Meituan’s implementation of on‑device deep learning models for search re‑ranking, covering the motivations for edge intelligence, feature engineering, feedback sequence modeling, model architecture, deployment optimizations, experimental results, and future directions, offering practical insights for developers building large‑scale AI on mobile.