DeepSeek hires 150 engineers to build the systems behind its AI agents

DeepSeek is hiring about 150 backend and server-side engineers as it expands its agent infrastructure. The roles cover systems beneath its models, including an elastic-computing platform for large numbers of sandbox environments.

DeepSeek agent infrastructure hiring round

The openings were posted across DeepSeek’s official recruitment portal in Beijing and Hangzhou. The roles include server development and Agent Elastic Compute research and development, with work spanning operating systems, virtual machines, storage, networking, scheduling and control-plane services. The New Stack reported that the hiring is focused on the backend systems supporting models and agents rather than a new model research push.

Unite.AI reported that the round was disclosed on September 7, citing a post by Tianyi Cui, a DeepSeek technical staff member. Cui said data volumes, machines, containers, training and evaluation tasks, agent environments, users and requests were all growing quickly, putting pressure on the company’s existing backend. The report said the company is looking to upgrade, maintain and rewrite those systems.

What DSec does for AI agents

DSec stands for DeepSeek Elastic Compute. DeepSeek’s technical report describes it as a production sandbox platform for agentic AI, with components for an API gateway, per-host agents and cluster monitoring. The paper says a single DSec cluster manages hundreds of thousands of concurrent sandbox instances in production.

That workload requires more than a model endpoint. Each agent environment needs compute, storage, networking and controls for starting, monitoring, recovering and stopping jobs. The systems work is less visible than a model launch, but it decides how many agents can run without the bill and failure rate getting out of hand.

Why the infrastructure push matters in the UAE

UAE companies building internal agents will face the same engineering questions when experiments move into production: how to isolate workloads, schedule thousands of jobs, keep data available and recover when a task fails. DeepSeek has not announced a UAE deployment or local hiring programme in the sources reviewed, so this is infrastructure context rather than a local product launch.

The timing also follows a wider expansion. Reuters reported in June that DeepSeek planned to at least double staff in every department, with roles including full-stack development, algorithms, product management and deep-learning research. The latest round puts more of that growth into DeepSeek agent infrastructure needed to operate agents at scale.