ModelBest
Posted 1w ago

法律智能算法实习生(大模型训练与数据合成方向)

ModelBest
Beijing, Beijing, China
OnsiteInternship
Responsibilities
  • training models
  • synthesizing data
  • evaluating models
Requirements
  • Participate in legal-domain large-model fine-tuning (SFT/DPO/RLHF)
  • Automated data synthesis
  • Reward-model construction and automated evaluation
  • Requires strong algorithms
  • Transformer and PyTorch experience
  • Internship ≥4 months, ≥4 days/week
Technical tools mentioned
PythonPyTorchLLaMA-FactoryTRLveRLDeepSpeed

Job description

1、法律大模型后训练:参与法律垂直领域大模型的 SFT / 偏好对齐(DPO / RLHF)训练全流程,围绕法律业务场景设计训练方案,优化模型在复杂约束下的生成质量与逻辑严密性;
2、法律训练数据合成:设计并实现自动化法律数据合成链路,利用 Multi-Agent 协同方案生成高质量 SFT / 偏好对齐数据;应用聚类与模式识别方法实现语料的深度清洗与多样化筛选,持续识别并优化合成链路的效率与质量瓶颈;
3、Reward Model 与自动化评测:参与构建针对法律文本逻辑严密性、条文引用准确性的奖励模型,应用 AI-as-Judge 技术搭建训练效果的自动化评测闭环,用评测结果反哺数据配比与训练策略迭代;
4、前沿技术探索:跟踪并复现 RL 训练(GRPO / PPO)、长思维链蒸馏、Agentic 任务训练等前沿工作,探索其在法律场景中的落地路径。

1、专业背景:计算机、人工智能或相关专业在读硕士/博士(优秀本科生亦可),具备扎实的数学与算法功底;
2、算法深度:深入理解 Transformer 架构与大模型训练原理,对 SFT / RLHF / DPO / PPO 等后训练技术有实战经验,完整跑通过至少一次大模型微调或对齐训练流程;
3、数据与工程能力:熟练使用 Python 与 PyTorch,具备大规模数据处理与 Pipeline 构建经验;熟悉 LLaMA-Factory / TRL / veRL / DeepSpeed 等训练框架者优先;有数据合成、数据配比消融实验经验者优先;
4、领域兴趣:对法律文本逻辑建模、复杂 Agent 系统有浓厚兴趣;有垂直领域模型训练、代码大模型或 Agent 训练项5、目经验者优先;有 RAG / 检索系统经验者加分(非必需);
6、学术背书:在 NeurIPS / ICLR / ACL / EMNLP 等顶级会议发表过大模型训练相关论文者优先;
7、实习要求:实习期不少于 4 个月,每周到岗不少于 4 天。

About ModelBest

Develops lightweight, on-device artificial intelligence models and agents.

Year founded
2022
Employees
200
Organization type
Private
Latest investment
Raised $738.00M Series C (2026)
Headquarters
CN

Similar jobs

Algorithm Intern roles near Beijing, Beijing
20h
Save
Mark Applied
Hide
算法实习生(蛋白质结构生成方向)
Beijing, Beijing, China
OnsiteInternship
北京科学智能研究院
北京科学智能研究院: A research institute focused on AI for Science innovation.
Current undergraduate or higher student in a relevant field; proficient in Python with PyTorch and JAX experience; interested in diffusion and generative models; able to implement research algorithms.
Python, PyTorch, JAX, AlphaFold3
3d
Save
Mark Applied
Hide
算法实习生 - 抖音
Beijing or Shenzhen
OnsiteInternship
ByteDance
ByteDance: Developing AI-driven content platforms and mobile applications.
Current bachelor's or higher student, preferably in computer science; machine learning foundation; Python/C++; deep learning framework experience; strong quantitative, self-learning, engineering, communication, and data analysis skills.
Diffusion Model, GPT, CV, NLP, Python, C++, PyTorch, TensorFlow, MXNet, GAN, VAE
4d
Save
Mark Applied
Hide
日常实习生-大模型算法(Agentic RL / Reward 方向)-Qwen基础模型
Beijing or Hangzhou or Shanghai
HybridInternship
Alibaba
AlibabaNYSE: BABA: Provides online marketplaces, cloud computing, and digital payment services.
Undergraduate or graduate student in computer science, AI, mathematics, or related fields; proficient in Python and PyTorch; strong machine learning and deep learning foundations; familiar with LLMs and interested in reinforcement learning.
Python, PyTorch, LangChain, LangGraph, Kaggle, PPO, GRPO
2w
Save
Mark Applied
Hide
算法实习生 - 运动规划
Beijing, Beijing, China
OnsiteInternship
Pony.ai
Pony.aiNASDAQ: PONY: Develops autonomous driving technology and operates robotaxi services.
Enrolled undergraduate or higher in computer-related field; strong data structures and algorithms; proficient in C++; able to commit ≥6 months and attend weekly; interest in autonomous driving.
C++, Linux, Git
2w
Save
Mark Applied
Hide
具身智能算法实习生
Beijing, Beijing, China
OnsiteInternship
Lingyu Intelligence
Lingyu Intelligence: Developing teleoperation systems and data infrastructure for embodied AI robots.
Seeking senior undergraduate or above in CS/Robotics/AI/Automation with Python and PyTorch skills, familiarity with RL/imitation learning, strong coding and engineering ability.
Python, PyTorch, MuJoCo, Isaac Sim, LeRobot
3w
Save
Mark Applied
Hide
用户理解算法实习生
Beijing or Shanghai
OnsiteInternship
Xiaohongshu
Xiaohongshu: Social media and e-commerce platform for lifestyle sharing.
Familiarity with machine learning/deep learning and data mining; proficient with at least one deep learning framework and one programming language; strong coding and communication skills; search/recommendation experience preferred.
tensorflow, pytorch, C++, python, java, scala, LLM, Agent
3w
Save
Mark Applied
Hide
Omni模型算法实习生
Beijing or Shenzhen
OnsiteInternship
XPeng
XPengNYSE, HKEX: XPEV, 9868: Designs and manufactures intelligent electric vehicles and mobility solutions.
Current senior undergraduate or master's student in CS/AI/automation/math/electronic information; able to intern ≥3 months with ≥4 days/week. Knowledge of Transformer/LLM/VLM, Python and PyTorch; experiment skills and strong documentation and collaboration.
Python, PyTorch, Transformer, LLM, VLM
4w
Save
Mark Applied
Hide
大模型应用算法实习生
Beijing, Beijing, China
OnsiteInternship
SenseTime
SenseTimeHKEX: 0020: Develops artificial intelligence software and computer vision technology.
Pursuing a bachelor+ in CS/EE/Statistics, strong fundamentals and project experience, Python proficiency, good communication; NLP/task experience preferred.
Python