Tagged articles
Ray
70 articles · Page 1 of 1
The article details how the Firefly AIGC platform evolved from a chat interface to a production-grade content generation system, covering its unified orchestration architecture, differentiated execution using Ray Serve and Ray Data, multimodal data lakehouse with Lance, and workflow observability via Trace.
AI infrastructureAIGCContent Production
0 likes · 27 min read


JD TechOct 29, 2026 · Artificial Intelligence
JD's Firefly platform shares its engineering practices for scaling AIGC from single model calls to a production-grade content generation system, covering unified orchestration with DAGs, differentiated execution via Ray Serve and Ray Data, multimodal data lakehouse with Lance, and workflow-level observability through Trace.
AI infrastructureAIGCContent Production
0 likes · 24 min read

JD Retail's Firefly AIGC platform moves beyond single-model generation to scalable content production by unifying orchestration, differentiated execution via Ray Serve and Ray Data, a Lance-based multimodal data lakehouse for material governance, and end-to-end traceability across DAG workflows.
AI InfraAIGCContent Production
0 likes · 22 min read

Benchmark tests show EMR Serverless Ray combined with Lance achieves sub-second vector retrieval on billion-scale autonomous driving datasets (768-dim, L2, top-20k), with P50 latency of 1.23s for pure vector search and 3.95s for multi-filter queries, enabling integrated multimodal Python workflows for embedding generation, reranking, and sample mining.
Autonomous DrivingEMR Serverless RayLance
0 likes · 31 min read

This article provides a comprehensive technical deep dive into Ray Data, covering its architecture, streaming execution model, shuffle implementation, fault tolerance, performance tuning, and best practices for AI data pipelines, positioning it as a superior alternative to Spark for GPU-intensive workloads.
AI Data PipelinesPerformance TuningRay
0 likes · 26 min read

This article introduces Daft, Ray, and Lance as the core layers of modern AI data infrastructure: Daft handles multimodal data computation, Ray manages distributed scheduling across heterogeneous resources, and Lance provides a lakehouse format optimized for vector search and random access, together replacing fragmented toolchains.
AI data infrastructureDaftLance
0 likes · 11 min read

Ray History Server solves the loss of observability when ephemeral Ray clusters terminate by archiving logs, events, and dashboard snapshots to object storage, enabling long-term querying, debugging, and unified access to both live and historical clusters via a reconstructed dashboard.
KubeRayKubernetesObject Storage
0 likes · 12 min read

The article reviews Tencent Cloud's AI DLC launch, detailing how the serverless Spark + Ray platform unifies data, compute, and agent workflows, introduces four architectural upgrades, showcases core engines (TCRay, Xpark, Meson, Open Engine), and presents benchmark results and real‑world practices from Bosch and WorkBuddy that demonstrate significant performance and productivity gains.
AI DLCRaySpark
0 likes · 14 min read

The article interviews three newly appointed core contributors—Ray community member Fan Shilun, Apache Auron committer Gui Hua‑wen, and Linux kernel maintainer reviewer Li Nan—who describe how they turned business pain points into open‑source contributions, the skills they gained, and actionable steps for engineers to start and grow in open‑source communities.
Apache AuronLinux kernelRay
0 likes · 9 min read

With agents moving from Q&A to execution, traditional architectures expose three critical flaws—data stored in lakes, models in the cloud, and split scheduling—forcing petabyte‑scale data movement; Tencent Cloud’s big data AI DLC resolves this by running Spark and Ray side‑by‑side on the same lake, enabling closed‑loop processing and automatic trajectory capture.
AIAgentRay
0 likes · 2 min read

The article analyzes how traditional data‑lake‑model‑cloud architectures expose three critical flaws for agentic AI—massive data movement, fragmented logging, and split compute—then details Tencent Cloud's Big Data AI DLC solution that unifies Spark and Ray on a single lake to enable in‑place processing, closed‑loop training, and cost‑effective iteration.
Agentic AIRaySpark
0 likes · 2 min read

The article explains how Tencent CodeBuddy’s AI DLC platform unifies Spark batch processing and Ray training to eliminate data movement, turning agent trajectories into reusable training fuel, which reduces monthly‑level training cycles to weekly, enables in‑place computation on billions of features, and cuts operational costs by 60%.
AI DLCGPU utilizationRay
0 likes · 2 min read

The article presents DataFlow, an open‑source, GPU‑centric data‑engineering framework that tackles LLM data‑preparation bottlenecks by defining a two‑level operator taxonomy, a LLM‑driven WebAgent for automatic crawling, a PDF‑to‑Markdown MinerU, a Ray‑based distributed runtime, and extensive multimodal extensions, and validates the design with quantitative experiments showing significant quality gains across math, code, and reasoning benchmarks.
DataflowLLMOperator
0 likes · 14 min read

Facing over 1 PB of daily sensor data, BYD replaced Airflow with a multi‑cluster Argo Workflows and Argo CD architecture, integrated Ray for GPU workloads, and achieved 20‑40 k concurrent workflows, an 11‑fold efficiency boost, 30% cost reduction, and near‑99% success rates.
Argo WorkflowsAutonomous DrivingKubernetes
0 likes · 11 min read

The article analyzes why 74% of AI projects fail due to feedback gaps and data silos, explains how the open‑source Lance format addresses these issues with unified multimodal storage, outlines a layered Lance‑on‑Ray architecture, and details three real‑world practices—implicit feedback loops, GPU‑accelerated self‑evolution, and semantic knowledge‑graph evolution—to boost R&D efficiency.
CAGRADaftGPU Indexing
0 likes · 13 min read

This article details a step‑by‑step, distributed pipeline built on Alibaba Cloud PAI using Data‑Juicer and Ray that transforms raw egocentric hand videos into LeRobot v2.0‑compatible Vision‑Language‑Action (VLA) training data, covering video splitting, frame extraction, camera calibration, 3D hand reconstruction, pose estimation, action captioning, and export, with code snippets, performance numbers, and references.
Data-JuicerEmbodied AILerobot
0 likes · 29 min read

When attribution analysis on over 900 million rows slowed to more than 40 seconds and threatened cluster stability, Ctrip's smart attribution team rebuilt the architecture with Ray and DuckDB, achieving sub‑15‑second query times, 160 % performance gain, and complete resource isolation.
DuckDBKubernetesRay
0 likes · 22 min read

This article details a step‑by‑step guide for constructing a high‑performance multimodal data pipeline—covering video segmentation, duration filtering, frame extraction, safety and aesthetic scoring, and caption generation—using Alibaba Cloud PAI, Paimon, DataJuicer, and distributed frameworks like Ray and Daft, with real‑world performance metrics.
AIAlibaba CloudDaft
0 likes · 30 min read

This article analyzes how the Daft‑Ray‑Lance stack tackles the challenges of multimodal AI workloads by offering a high‑performance Rust engine, adaptive back‑pressure, seamless Ray‑based distributed scheduling, and a storage format optimized for random access, vector indexing, and zero‑copy schema evolution, complete with benchmark comparisons and practical deployment guidance.
DaftData EngineeringLance
0 likes · 21 min read

This guide explains the end‑to‑end workflow for constructing a distributed scalar index in Lance by orchestrating validation, fragment sharding, worker‑level indexing via Ray, and final metadata merging, complete with code snippets and detailed step‑by‑step instructions.
LancePythonRay
0 likes · 12 min read

This article details Ant Group's use of the Ray distributed computing framework to accelerate massive data indexing, migrate a C++ engine to Ray, implement elastic resource scheduling, improve long‑tail task efficiency, and build a robust RAG operator system with comprehensive governance, achieving up to 2× speed gains and 99.9% success rates.
Rayai data pipelinebackend development
0 likes · 15 min read

This article explains the evolution of Data‑Juicer from a pure‑text preprocessing tool to a full‑stack multi‑modal data engine, detailing its architecture, operator library, Ray‑based distributed execution, performance benchmarks, integration with AI agents, and roadmap for future AI‑centric data workflows.
Data-JuicerMulti-modalRay
0 likes · 31 min read

This article provides a comprehensive, step‑by‑step explanation of Ray Data's LogicalPlan architecture, covering its class hierarchy, core methods, logical operators, optimization rules, planning from logical to physical operators, execution binding, metadata inference, lineage serialization, and the full file/module index for developers building scalable data pipelines.
BackendLogicalPlanRay
0 likes · 35 min read

This comprehensive guide explains Ray Data's Dataset core type, its distributed pipeline design, lazy execution model, API groups, transformation methods, column operations, I/O integrations, metadata utilities, and execution workflow, providing clear code examples and practical usage tips.
APIDatasetDistributed Data
0 likes · 18 min read

Python’s reputation for slowness can be overcome by selecting the right tools—Numba, PyPy, CuPy, JAX, Ray, Joblib, async I/O, memory profilers, and big‑data frameworks—delivering speedups from 6× to over 50× with minimal or no code modifications.
GPURayasync
0 likes · 22 min read


JD TechJan 31, 2026 · Artificial Intelligence
This article details JD Retail's 9N‑LLM unified training framework that tackles the massive data, hardware heterogeneity, and algorithmic challenges of generative recommendation by integrating TensorFlow and PyTorch, supporting GPU/NPU, and delivering high‑throughput sample processing, sparse/dense optimization, and flexible reinforcement‑learning capabilities.
GPU/NPURaylarge-scale AI
0 likes · 26 min read

This article examines the shortcomings of traditional big‑data engines for AI workloads, presents a Ray‑based heterogeneous fusion architecture that unifies CPU/GPU scheduling, Python ecosystems, and streaming‑batch processing, and details fault‑tolerance, checkpointing, compute‑storage separation, resource‑utilization, scalability, and observability improvements that enable thousands of nodes and dramatically higher GPU efficiency.
RayResource Optimizationai data pipeline
0 likes · 31 min read

This article details a scalable, distributed pipeline that uses LAS AI Data Lake, Daft on Ray, and advanced video‑processing techniques—scene detection, splitting, frame sampling, filtering, and caption generation—to transform tens of millions of hours of robot‑captured video into high‑quality, searchable semantic data while dramatically boosting CPU and GPU utilization.
AI PipelineDaftRay
0 likes · 21 min read

This guide explains Ray's fundamental building blocks—including Tasks, Actors, remote Objects, Placement Groups, and environment dependencies—showing how to define, schedule, and retrieve distributed workloads with code examples and command‑line utilities.
ActorsObject StoreRay
0 likes · 8 min read

This article examines how WeChat’s Astra platform leverages the Ray distributed framework to manage million‑node AI workloads, addressing challenges of scale, heterogeneous GPU resources, operational complexity, and cost, and outlines the architecture that unifies Ray services across multiple Kubernetes clusters.
AI scalingAstra PlatformGPU Management
0 likes · 5 min read

This article examines how Tencent's WeChat team leveraged the Ray distributed computing framework within the Astra platform to tackle massive AI workloads, addressing challenges of scale, GPU diversity, operational complexity, and cost while outlining their architecture and practical insights.
AI infrastructureAstra PlatformRay
0 likes · 6 min read

This article details how Pinterest’s senior staff engineer Dr. Luo leveraged the open‑source Ray framework to overcome LLM data‑preprocessing bottlenecks, describing the system’s architecture, key features such as map_batches, Carry‑Over Columns and Accumulators, and the dramatic performance and cost improvements achieved.
LLMPinterestRay
0 likes · 12 min read

This article examines how Tencent leverages the Ray distributed framework within the Astra platform to handle WeChat's massive AI workloads, addressing challenges of scale, heterogeneous GPU resources, operational complexity, and cost while outlining the architecture and practical benefits.
AI scalingAstra PlatformRay
0 likes · 5 min read

The article presents Ant Group’s Ray‑based Ragent framework, detailing its background, motivation behind unified AI serving, and the four core modules—Profile, Memory, Planning, and Action—that together enable large‑language‑model agents for financial applications.
AI frameworkAnt GroupLLM Agents
0 likes · 4 min read

This article introduces Ant Group’s Ragent, a Ray‑based distributed AI agent framework, covering its background, motivation in the large‑model era, and a four‑module design (Profile, Memory, Planning, Action) that enables scalable LLM‑driven agents.
AI frameworkAnt GroupLLM Agents
0 likes · 4 min read

This article introduces Ant Group’s new Ray‑based distributed agent framework Ragent, outlines its background and motivation, and details the four core modules—Profile, Memory, Planning, and Action—that together enable sophisticated LLM‑driven AI agents for large‑scale applications.
AI agentsAnt GroupLLM
0 likes · 4 min read

This article introduces Ant Group’s Ragent, a Ray‑based distributed AI‑agent framework, covering its background, motivation, design and implementation, and detailing the four core modules—Profile, Memory, Planning, and Action—that enable large‑language‑model agents at massive scale.
AI agentsAnt GroupRagent
0 likes · 4 min read

This article introduces Ant Group’s Ragent, a Ray‑based distributed AI Agent framework, detailing its background, motivations, and design, and explains the four essential modules—Profile, Memory, Planning, and Action—that enable scalable large‑language‑model agents for real‑world applications.
Ant GroupRay
0 likes · 5 min read

This article introduces Ant Group’s Ragent, a Ray‑based distributed agent framework, covering its background, motivation, and design—including core modules like Profile, Memory, Planning, and Action—that enable large‑language‑model agents to operate at massive scale.
Ant GroupRagentRay
0 likes · 5 min read

This article introduces Ant Group’s Ray‑based distributed agent framework Ragent, outlines its background, motivation, and design, and details the four essential modules—Profile, Memory, Planning, and Action—that power large‑language‑model agents in large‑scale AI serving.
AI agentsAnt GroupLLM
0 likes · 5 min read

This article explains how Ant Group’s Ragent framework leverages Ray to create scalable, multi‑tenant AI agents, detailing its background, motivation, and design while outlining the core modules—Profile, Memory, Planning, and Action—that power large‑language‑model agents.
Ant GroupRaydistributed computing
0 likes · 5 min read

This article explores how the LAS team's AI‑driven data lake solution, built on Daft for lake computing and Lance for lake storage, tackles the emerging challenges of multimodal data handling, offering faster I/O, heterogeneous CPU‑GPU scheduling, and seamless integration for AI workloads.
AIDaftLance
0 likes · 11 min read

This article introduces Ant Group’s Ray‑based distributed agent framework Ragent, outlines its background, motivation, and design, and breaks down the four essential modules—Profile, Memory, Planning, and Action—that enable large‑language‑model agents to operate in real‑world scenarios.
Ant GroupLLMRagent
0 likes · 5 min read

This article explains how Tencent's WeChat team built the Astra platform on Ray to manage millions of AI compute nodes, addressing challenges of massive scale, heterogeneous GPU resources, low‑priority node instability, deployment complexity, and cost, while detailing architecture, scheduling strategies, and practical usage examples.
AI scalingRayWeChat
0 likes · 21 min read

In the era of data‑driven AI, Ray offers an open‑source unified compute framework that abstracts distributed system complexity, enabling developers to seamlessly scale Python code from a laptop to large GPU clusters, and provides the Ray AI Runtime (AIR) with libraries such as Ray Data, Train, Tune, and Serve to accelerate LLM training, hyper‑parameter tuning, and model serving.
AI RuntimeLLM trainingPython
0 likes · 10 min read

This guide details comprehensive security best practices for deploying Ray clusters on Alibaba Cloud ACK, covering TLS communication, namespace isolation, resource quotas, RBAC, security contexts, image scanning, resource limits, RRSA integration, multi‑cluster isolation, and recommendations for protecting dashboards and services from unauthorized access.
ACKKubernetesRay
0 likes · 18 min read

This article introduces the integration of the open‑source distributed AI framework Ray with Alibaba Cloud’s PAI platform, detailing its advantages, architecture, fault‑tolerance, resource management, and provides a comprehensive step‑by‑step tutorial—including configuration, command examples, and code snippets—to efficiently run Ray jobs on PAI‑DLC.
Alibaba CloudPAIRay
0 likes · 11 min read

This article explains how to build a flexible machine‑learning infrastructure on Alibaba Cloud ACK using Ray and KubeRay, covering Ray's core components, AI libraries, deployment options on VMs and Kubernetes, code examples for data processing, model serving, and advanced scheduling and quota management techniques.
AIAlibaba CloudKubeRay
0 likes · 17 min read

This article explains how an asynchronous RLHF pipeline built on vLLM, Ray, and OpenRLHF dramatically reduces training bottlenecks by decoupling inference, environment interaction, and model updates, and provides detailed implementation code and design choices for scalable reinforcement learning.
OpenRLHFRLHFRay
0 likes · 11 min read

This article explores the emerging concept of agentic reinforcement learning for large language models, analyzes ByteDance's VeRL and the Search‑R1 frameworks, identifies practical challenges in tool integration and environment parallelism, and proposes a unified, Ray‑based architecture to enable scalable, high‑quality RL environments.
RaySearch-R1environment design
0 likes · 11 min read


AntDataApr 3, 2025 · Artificial Intelligence
Ray Flow Insight is an Ant Group open‑source tool that visualizes Ray's distributed programming primitives—Actors, Tasks, and Objects—to turn complex reinforcement‑learning systems from opaque "black boxes" into transparent, debuggable workflows, providing logical, physical, distributed stack, and flame‑graph views for performance analysis and optimization.
AIDebuggingRay
0 likes · 32 min read

This article introduces DeepSeek Smallpond, a lightweight yet high‑performance AI data‑processing engine built on Ray and DuckDB, explains its dual Dataframe and LogicalPlan APIs, showcases integration with Volcano Engine's AI Data Lake LAS, and provides practical code examples for distributed processing, multimodal storage, and RAG pipelines.
AI data processingDuckDBLance
0 likes · 18 min read

This article introduces Ragent, Ant Group’s Ray‑powered distributed agent framework, covering its background, motivation, design, implementation details, multi‑agent capabilities, and future directions for large‑model AI applications.
AIDistributed AgentsPython
0 likes · 14 min read

The article details ByteDance's use of Ray and RayData to construct scalable audio and video data processing pipelines for multimodal AI models, addressing challenges of massive data volume, resource constraints, and fault tolerance through pipeline design, RayCore enhancements, and custom scheduling optimizations.
AIByteDanceMultimodal Models
0 likes · 16 min read

eBay upgraded its AI infrastructure by adopting Ray, cutting model development and deployment time by roughly 50% and boosting GPU utilization from about 10% to over 75% through automated cluster scaling and high‑throughput batch inference.
AI infrastructureData+AIGPU utilization
0 likes · 5 min read

This article details the AI computing challenges faced by WeChat, explains why the Ray distributed engine was chosen, and describes the design and large‑scale deployment of the AstraRay platform—including scheduling, resource management, and multi‑model support—to achieve low‑cost, high‑efficiency AI services.
AI PlatformAstraRayRay
0 likes · 20 min read

This article walks through a common Modin pandas Excel import error, explains why installing an execution engine like Ray or Dask is required, and shows how renaming a conflicting local file resolves the ModuleNotFoundError, enabling successful data loading.
Raydaskmodin
0 likes · 4 min read


Didi TechJan 25, 2024 · Artificial Intelligence
Didi’s new Ray‑native XGBoost training platform replaces the fault‑prone Spark solution with a fully Pythonic, fault‑tolerant architecture that leverages Ray’s autoscaling and gang‑scheduling, delivering 2–6× speedups, reduced failure rates, efficient sparse‑vector handling, scalable hyper‑parameter search, and improved resource utilization for large‑scale machine‑learning workloads.
MLOpsRayXGBoost
0 likes · 20 min read

This article explains why Ray is popular among AI researchers, how ByteDance uses KubeRay to host Ray applications, and how Kueue manages and schedules RayJob workloads, covering Ray's architecture, KubeRay components, real-world use cases, and job scheduling strategies.
AIKubeRayKubernetes
0 likes · 12 min read

This article explains how the Ray distributed computing engine simplifies the design, deployment, and operation of complex cloud‑native distributed systems—illustrated through an AutoML service example—by detailing system complexity, Ray’s core concepts, resource customization, runtime environments, monitoring, and ecosystem integrations.
AIAutoMLKubernetes
0 likes · 26 min read


AntTechJun 27, 2023 · Artificial Intelligence
Fanglue is an interactive, web‑based rule‑development platform that integrates expert domain knowledge with distributed AI algorithms to efficiently generate and evaluate decision rules for anti‑fraud scenarios, leveraging Ray for real‑time processing and achieving VLDB‑2023 acceptance.
AIRayVLDB2023
0 likes · 10 min read

Large-model offline (batch) inference, which processes massive data on billion-parameter models, faces GPU memory and distributed scheduling challenges; this article explains how Ray's cloud-native framework, model parallelism, and Ray Datasets pipelines address these issues, improve throughput, and enable elastic, efficient GPU utilization.
GPU utilizationLarge Model InferenceRay
0 likes · 16 min read

This article explains the challenges of large‑model offline (batch) inference, such as GPU memory limits and distributed scheduling, and shows how Ray’s cloud‑native architecture, model partitioning, and Ray Datasets can be used to build efficient, elastic inference frameworks deployed with KubeRay.
GPU memoryRaydistributed computing
0 likes · 18 min read


AntTechJan 3, 2023 · Artificial Intelligence
Ray, an open‑source distributed computing framework originally created by Berkeley's RiseLab and heavily contributed to by Ant Group, underpins many AI workloads—from privacy‑preserving federated learning to large‑scale model training for ChatGPT—making it a critical yet often overlooked engine of the generative AI revolution.
OpenAIRaymachine learning
0 likes · 7 min read


Code DAODec 17, 2021 · Artificial Intelligence
The article explains why XGBoost training can be slow despite its speed focus and presents three acceleration techniques—choosing an optimal tree_method, leveraging cloud resources for larger memory, and using Ray for distributed training—complete with code examples and benchmark results.
RayXGBoostcloud computing
0 likes · 5 min read


Code DAODec 17, 2021 · Artificial Intelligence
XGBoost‑Ray provides a fault‑tolerant, multi‑node, multi‑GPU backend for XGBoost that integrates seamlessly with Ray Tune, supports distributed data loading, and can be enabled with only three code changes, enabling scalable training and inference on large clusters.
GPURayRay Tune
0 likes · 8 min read

This article traces the history of big‑data computing engines—from early MapReduce and Hadoop through Spark, Storm, Flink, and the newer Ray—explaining their technical advances, real‑world applications in AI and finance, and why graduates should consider a career in this rapidly evolving field.
AICareerRay
0 likes · 16 min read


AntTechMar 1, 2021 · Artificial Intelligence
The article explains how Ant Group tackles the challenge of tightly integrating multiple computing paradigms by building a Ray‑based fusion engine, detailing its architecture, features, large‑scale applications in online machine learning and parallel processing, and outlining future development and recruitment opportunities.
Ant GroupFusion EngineRay
0 likes · 10 min read


AntTechDec 4, 2019 · Artificial Intelligence
The interview details how Ant Financial transitioned from offline to online machine learning by adopting the Ray distributed engine, describing their open architecture, fusion computing approach, technical advantages, encountered pitfalls, and plans to open‑source the system for broader AI and big‑data use.
AIAnt FinancialRay
0 likes · 15 min read


AntTechNov 1, 2019 · Artificial Intelligence
The article describes how Alipay tackled the cold‑start, conversion, and user‑experience challenges of its time‑limited “Collect Five Blessings” activity by designing a unified online machine‑learning system based on the Ray distributed‑computing framework, emphasizing stability, efficiency, simplicity, multi‑language support, and fault‑tolerant scheduling.
AlipayRayonline machine learning
0 likes · 11 min read
