- The hiring push focuses entirely on backend development and elastic computing for AI agents
- The shift suggests the Chinese AI startup is putting more weight on infrastructure as its models, users and computing demands scale
DeepSeek has opened around 150 engineering positions in its latest hiring push, with not a single role dedicated to AI research — a notable departure for a company better known for its breakthroughs in large language models.
The positions, announced September 7, are concentrated in two areas: backend development and agent elastic-computing research and development.
A departure from previous hiring campaigns
Most target engineers with two to 10 years of experience, rather than the young researchers and fresh graduates DeepSeek has traditionally favored.
If filled, the 150 positions could expand DeepSeek’s estimated 300-to-500-person team by roughly a third or more.
“Anything in computing becomes dramatically more complex as it grows in scale,” Cui Tianyi, head of DeepSeek’s Harness team, said.
He pointed to rapidly growing volumes of data, machines and containers, training and evaluation tasks, agent environments, users and requests.
Focus on DSec
The hiring underscores DeepSeek’s growing focus on its in-house elastic computing platform, DSec, or DeepSeek Elastic Compute.
First disclosed in the company’s DeepSeek-V4 technical report, the platform is built from three Rust components covering the API gateway, host-level Edge agent and cluster monitor.
In effect, the system is designed to keep DeepSeek’s growing AI workloads running reliably and efficiently across large numbers of machines.
According to the recruitment postings, DSec will require changes across the entire technology stack, “from the operating system to virtual machines, and then to networks and storage at all levels.”
The postings also acknowledge that many of the problems have no established solutions to follow.
A broader challenge
The shift reflects a broader challenge facing AI companies as models move from research labs into large-scale deployment.
More data, machines, training and evaluation tasks, agent environments, users and requests all add layers of complexity that cannot be solved by model improvements alone.
For DeepSeek, the latest hiring spree suggests that scaling AI is increasingly becoming an infrastructure problem — and one the company wants to solve largely with its own engineering capabilities.
