Agent Harness Researcher (Intern/Full-time)

Agent Harness Researcher (Intern/Full-time)

Share

Full-time |Full-stack development and algorithmsDeepSeek|Zhejiang · Hangzhou City, Beijing City

Apply for a position

Job description

Team mission

Model + Harness = Agent

We are transforming DeepSeek's cutting-edge model capabilities into leading agent products. All work except the model itself falls under the scope of Harness.

You will join the Harness team, working closely with researchers, engineers, and product managers to explore the research frontiers in the Harness field and define DeepSeek's understanding of Harness.

Main responsibilities

• Researchers and engineers from the Harness team communicate deeply and work closely to jointly define and realize cutting-edge model-capability-based innovations in the Harness field, including but not limited to context management, long-term memory, Subagents and Multi-Agents, and auto-evolving Agents.

• Deep communication and collaboration with researchers and engineers from the model training team enable co-evolution of models and Harnesses, achieving deep adaptation between DeepSeek's Harness and models from the Harness perspective.

• Benchmarking and evaluation methods for the Harness domain are proposed, benchmarking data is constructed, and data annotation strategies are formulated, aiming to study and optimize the intelligence level of agents in various domains from the Harness perspective.

• Using real-world tasks as an important feedback source for Harness research, designing relevant data and experiments to continuously iterate the Agent's performance in real-world usage scenarios.

• Based on user feedback collected by the team, we studied and optimized Harness's ability to solve real-world problems from the broadest possible user base.

Job Requirements

• Over 2 years of research experience in computer science or related fields, with strong skills, broad vision, and a strong sense of scientific research. Exceptionally outstanding candidates may have their years of service extended for them.

• Master's degree or above. Exceptionally outstanding candidates may be relaxed.

• Published valuable papers in the field of computer science.

• Able to independently analyze and propose their own ideas when facing problems, and possess the ability to drive research from zero to one. Able to quickly turn ideas into actionable prototypes, with efficient experimental iteration capabilities.

• Proficient in software development using AI Agent tools. Strong learning ability in the field of software development. Able to conduct research programming tasks in fields without direct experience (such as languages, technology, frameworks, etc.) with AI assistance.

• A high-intensity user of Agent products, with great enthusiasm for Agent Harness development and research, and a taste and judgment in model behavior. Deeply used code-based and general-purpose agent products, integrating related products into their work and daily life.

• Familiar with the basic mechanisms and technical principles of LLMs and agents, including LLM API, KV Cache, Agent Loop, Tool Use, Reasoning, Planning, Skills, MCP, Memory, Subagent, Multi-Agent, and other related knowledge. In-depth understanding of topics such as Prompt Engineering, Context Engineering, and Harness Engineering.

• Good Chinese communication skills.

Bonus items

• Research experience in AI-related fields, as well as research experience in the AI industry.

• Personal open-source works and in-depth contributions to the open-source community.

• Prestigious competition awards in the fields of computer science, programming, or data science.

• Other extraordinary talents related to this job.

Job information

Job title

Agent Harness Researcher (Intern/Full-time)

Salary range

-

Nature of the position

Full-time

Types of functions

Full-stack development and algorithms

Department

DeepSeek

Work location

Zhejiang · Hangzhou City, Beijing City