ML / AI Jobs in China

运筹调度算法专家

Open 18d APAC

Senior operations-research/scheduling algorithm expert at Meituan in Beijing: builds and optimizes dispatch/scheduling algorithms for real business operations, monitors and improves them using data analysis and simulation, and models problems in collaboration with other teams. Core stack includes OR models plus Java, Python, or C++.

Python

Meituan

LongCat - 大模型基座认知分析算法研究员（多模态评测方向）

New APAC

Beijing, Shanghai

Researcher on Meituan's LongCat foundation-model team analyzing multimodal LLM capabilities (image/video/speech understanding and generation, embodied intelligence/VLA, GUI agents) and how training data and strategies affect model ability, producing evaluation schemes and training improvements. Core stack: Python/C++, deep learning, multimodal large models.

Meituan

Keeta技术-大模型算法工程师/专家

Open 18d APAC

AI Java Python PyTorch RAG +2 skills

LLM algorithm engineer/expert for Meituan's international Keeta food-delivery platform: builds intelligent customer service, content moderation, and liability-adjudication systems, including post-training base models (SFT/RLHF/DPO). Core stack: Python/Java, PyTorch/TensorFlow, RAG, agents, and multimodal models for multilingual markets.

Tencent

微信小程序-商业化算法工程师

New APAC

Guangzhou

渴望为世界带来新意的人，早已对描绘未来的“热词”不陌生——人工智能、机器学习、数据科学等等。腾讯始终在业界前沿不断探索，积极布局，并为优秀的你提供无限广阔的舞台。我们坚信科技的不断革新，最终能带来美好的、撼动人心的改变。 1.负责微信小程序生态内的商业化相关算法研发，包含但不限于商业化内容（短剧等）相关目标的召回和精排算法、变现策略模型的优化迭代；…

C++

State Street

Artificial Intelligence & Machine Learning, AVP

New APACFull-time

Agentic AI AI API AutoGen AWS +34 skills

Hangzhou

A highly experienced, hands-on technology leader to serve as Principal Agentic AI Engineer. This role will report to the AI Platform leadership team and will define the target-state Agentic AI platform architecture,…

SHEIN

反爬风控算法工程师-上海南京

New APACFull-time

Nanjing, Shanghai

岗位职责 1. 构建反爬"感知-识别-评估"体系。建设爬虫数据采集-特征-模型能力。 2. 跟踪常见爬虫技术与对抗手段，熟练掌握爬虫检测的各种方法。 3. 建立反爬风险水位和反爬价值评估方法。 4. 和策略团队一起推动反爬能力体系建设,沉淀方法论与知识库。 5. 了解LLM和Agent在反爬领域的应用。任职要求 - 硕士及以上学历,计算机、数学、统计、信息安全等相关专业优先。 - 3年以上反爬、设备风险相关经验。 - 熟练使用…

Python SQL

千问事业部-深度学习应用开发工程师-杭州

Hangzhou

1. 负责夸克智能视觉相关业务服务，负责深度学习算法服务的流程设计及研发工作 2. 深入理解业务（扫描滤镜、文字服务、图像编辑、图像生成等），和算法紧密合作，对已有服务进行全链路的改进和优化 3. 技术预研和技术难点攻关，引入业界新技术和系统化方法，提升服务迭代效率，保障服务的稳定性、高性能和可扩展性 1. 本科及以上学历，计算机、软件工程相关专业优先 2.…

Open 17d APACSenior H-1B sponsor history

高德-高级定位算法工程师-定位

1、负责车道级定位核心算法研发，基于 ESKF、多传感器融合及规则引擎/端到端模型，构建高鲁棒性的定位系统。 2、设计并实现 GNSS、IMU、视觉、地图等多模态信息融合方案，优化复杂场景（高架、隧道、城区峡谷、岔路口等）下的定位稳定性与连续性。 3、负责车道匹配、拓扑关联、时序跟踪等核心模块开发，提升定位精度、稳定性与异常恢复能力。 4、探索端到端定位与时序建模技术，结合…

C++ Python

Token Foundry-大模型Post-training算法研究员-Qwen

Research role on Alibaba's Qwen (Tongyi Lab) large-model post-training team: improving LLMs after pretraining using RL, SFT and RFT — scalable verifier signals, reward models, reasoning-path compression, and agent/multimodal reasoning. Core stack: Python, PyTorch, LLM training; based in Beijing, Hangzhou or Shanghai.

AI Python PyTorch SGLang

千问事业部-大模型应用算法工程师（Agent 方向）-北京

Alibaba's Qwen business unit seeks an LLM application algorithm engineer to build autonomous agents for its advertising platform: agent planning/execution loops, memory and self-evolution systems, agent harness infrastructure (context engineering, multi-agent orchestration, evaluation, HITL), plus post-training (SFT/RL). Core stack: Python and agent frameworks such as Agno.

Python

Token Foundry-RLHF算法研究员-Qwen

AI DeepSpeed LLM NLP Python +2 skills

RLHF algorithm expert on Alibaba's Qwen team (Token Foundry): builds user-preference data pipelines, trains Reward Models, mines Badcases from user feedback, and develops LLM-based judge/verifier systems to align Qwen LLMs, using Python, PyTorch, and RL methods such as PPO, GRPO, and DPO.

Token Foundry-合版算法研究员-Qwen

DeepSpeed LLM Python PyTorch SGLang +1 skills

Researches model-merging (合版) algorithms for Alibaba's Qwen LLMs — on-policy distillation, knowledge fusion and SFT objectives to combine multiple capabilities into one model with minimal conflict/forgetting — plus data-mixing strategy, automated multi-capability evaluation pipelines, and end-to-end training/eval automation. Core stack: Python, PyTorch, LLM post-training.

Token Foundry-代码大模型算法研究员-Qwen

Researcher on Alibaba Token Foundry's Qwen code-focused LLMs: builds Qwen and Qwen-Coder through data collection, large-scale pre-training, RL post-training, and evaluation, and develops coding agents and computer-use (GUI) agents. Requires an MS/PhD in CS/AI/ML with deep learning and LLM research experience.

Token Foundry-大模型评估系统算法研究员-Qwen

Algorithm researcher on Alibaba's Qwen 'Token Foundry' team building LLM evaluation systems: constructing eval datasets that probe model weaknesses, training LLM-as-a-Judge and reward models to capture human preference, designing reward signals with RL to raise model capability, and building the engineering frameworks for eval/reward pipelines. Core stack: Python, PyTorch, Transformers/LLMs.

LLM Python PyTorch

Token Foundry-Reward System算法研究员-Qwen

Algorithm engineer on Alibaba's Qwen team designing, training, and iterating the Agentic Reward System for the Qwen foundation model — building reward signals for coding/agent scenarios (reward models, code verifiers, user-experience rewards) and reward infra coupled with CPT/SFT/RL pipelines (PPO/GRPO) to improve alignment quality.

Agentic AI

Token Foundry-RL算法研究员-Qwen

RL Algorithm Researcher on Alibaba's Token Foundry team advancing the Qwen foundation models: researching PPO/GRPO large-scale RL training, building Agentic RL pipelines (reward models, critics, verifiers), and iterating training data and environments for coding and long-horizon agents to improve reasoning, planning, and code generation.

Agentic AI

Token Foundry-网络安全大模型算法研究员-Qwen

Researcher on Alibaba's Qwen team improving large language model performance on cybersecurity tasks through post-training and reinforcement learning. Day to day: building security training/evaluation environments (cyber ranges, vulnerability reproduction, patch validation, attack-defense drills) and applying LLM/RL/Agent methods with PyTorch.

Token Foundry-预训练模型评测算法工程师-Qwen基础模型

Hangzhou

Join Alibaba's Token Foundry team building evaluation for the Qwen foundation models: design scalable automated LLM/agent benchmarks and datasets, tune pretraining + light-SFT data mixes, and mine metrics that predict a pretrained base model's post-training potential. Core work spans LLM evaluation, benchmark/data synthesis, and pipelines.

Token Foundry-大模型预训练算法研究员-Qwen

LLM pre-training algorithm engineer on Alibaba's Qwen team (Token Foundry), researching pre-training data synthesis and optimization, training strategies (loss functions, scaling laws, optimizers), and model architecture (MoE, attention variants, KV-cache compression, decoding speedup) to advance foundation models. Open to MS/PhD graduates with hands-on research experience.

CUDA Triton

Token Foundry-训练系统算法研究员-Qwen