Chinese AI start-up DeepSeek has begun hiring interns to process medical data as hospitals have begun adopting the company’s open-source models for use in clinical settings.

The company is offering 500 yuan ($70, £50) a day for interns who can work four days a week in Beijing to label medical data for applications involving advanced auxiliary diagnosis tools, according to ads posted on the Boss Zhipin jobs website, local media reported.

Applicants should have a medical background and either be undergraduates in their fourth year or have a master’s degree, with experience in using large language models (LLMs) and the ability to write Python code and LLM prompts, the ads say.

Clinical use

The job descriptions posted online, which were not listed on DeepSeek’s official WeChat hiring channel, said applicants would need to be able to design models for medical questions and create evaluation processes for the model’s medical capabilities.

The notice says the interns’ work is intended to help improve DeepSeek’s medical capabilities, including improving their medical knowledge and reducing false output in medical questions and answers.

Generative AI systems are known for creating results that appear plausible, but can be subtly distorted or completely fabricated – a particularly critical issue where it comes to clinical use.

The ads are the first time DeepSeek has publicly mentioned the need for medical data, but the company’s models, which came to worldwide notice earlier this year for their low cost and high performance, are already in clinical use.

At least 300 hospitals in China have begun using DeepSeek’s LLMs in clinical diagnostics and medical decision support as of March, the South China Morning Post reported.

In a May paper published in the medical journal JAMA, a team of researchers including the founding head of the prestigious medical group Tsinghua Medicine, Wong Tien Yin, warned that the trend could be dangerous.

DeepSeek’s tendency to generate “plausible but factually incorrect outputs” could lead to “substantial clinical risk” and could also pose privacy issues, the paper said.

R1 upgrade

DeepSeek said a significant upgrade to its R1 reasoning model released this month has given it performance comparable with the latest Western models from the likes of OpenAI and Google and also achieved a 50 percent reduction in hallucinations.

It said the R1-0528 release, the first major update to the R1 “reasoning” model released in January, beats the performance of Alibaba’s Qwen3, which was released a month earlier.

The improved capabilities were largely the result of adding more computing resources at the post-training stage, the DeepSeek said.

Benchmark tests including maths, coding and general logic showed that R1-0528 surpassed domestic Chinese models and attained similar levels to global models including OpenAI’s o3 and Google’s Gemini2.5-Pro, the company said.