# DeepSeek Hiring Team to Build Agentic AI Competitor to Claude Code
DeepSeek has posted 17 job openings fully focused on developing agentic AI. The positions include deep learning algorithm researchers for agents, experts in evaluating agent data, and infrastructure engineers. Candidates must have experience applying DeepSeek models in search scenarios, content generation, multimodal processing, and agent tasks. Requirements emphasize a deep understanding of reinforcement learning (RL) techniques, where models adjust strategies based on analyzing action outcomes.
Key emphasis in the descriptions — superiority over Anthropic's Claude Code and Cursor. This signals direct competition with market leaders in AI-based coding tools.
Context of Agentic AI Development in China
The hiring coincides with growing interest in agent systems in China. OpenClaw, an open-source AI agent, is integrating into WeChat and other messengers, driving mass adoption. Over 600 million users in the country already use generative AI. Major companies — Tencent, Alibaba, Baidu, ByteDance — are actively launching products based on OpenClaw.
DeepSeek is shifting from base models to full-fledged agent products, integrating them into real-world applications. This is a logical step for scaling expertise in RL and agent architecture.
- Key positions in hiring:
- Deep learning algorithm researcher for agents.
- Expert in evaluating and annotating agent data.
- Agent infrastructure engineer.
- RLHF specialist (Reinforcement Learning from Human Feedback).
- Technical requirements:
- Experience with DeepSeek models in agent scenarios.
- Knowledge of RL techniques for iterative agent improvement.
- Competitor analysis: Claude Code, Cursor.
Parallel Work on DeepSeek V4
The company isn't stopping at agents: DeepSeek V4 release is expected, optimized for coding and long context. The model was trained on Huawei and Cambricon chips, without Nvidia. This strengthens independence from Western technologies.
DeepSeek is building a full stack: from hardware to agent tools. Infrastructure on domestic chips allows scaling model training tailored to agent tasks, which demand high efficiency on long sequences.
What’s Important
- DeepSeek focuses on 17 specialized roles for agentic AI, explicitly targeting superiority over Claude Code and Cursor.
- Growth of agent systems in China: OpenClaw integrates into messengers, used by 600+ million people.
- DeepSeek V4 — first model on Huawei/Cambricon chips, optimized for coding and long context.
- Strategy: full stack from hardware to agents, with emphasis on RL and independence from Nvidia.
- For developers: positions require deep knowledge of RL in the context of DeepSeek models.
— Editorial Team
No comments yet.