Should Graduate Students Choose Embedded Control or Machine Vision?

Machine Vision: Competition and Uncertainty

AI‑related majors have been expanded rapidly, leading to a large supply of graduates who can run YOLO models and answer Transformer questions. The market becomes saturated within three years of graduation, and the technology cycle is extremely fast—today’s model architecture can be replaced by a new state‑of‑the‑art method tomorrow.

High‑quality algorithm positions typically require top‑conference publications or internships at major technology companies. Without such credentials, even strong 985 master’s graduates are likely to be limited to parameter‑tuning and experimental work rather than full‑stack AI engineering.

Work intensity in visual‑algorithm roles is reported as 996 (9 am–9 pm, six days a week) as a norm, with project crunches extending to 007, making long‑term sustainability difficult.

Embedded Systems: A Stable, In‑Demand Skill Set

Embedded engineering is required for every hardware product—smart‑home devices, industrial control, automotive electronics, medical equipment—so demand does not depend on AI hype and is expected to grow with the Internet of Things.

The core technology stack (C, RTOS, communication protocols) has remained largely unchanged for a decade, reducing the need for continual re‑learning of new research papers.

Entry‑level salaries for fresh master’s graduates at companies such as Huawei, DJI, and BYD are typically 20 – 25 wan RMB per year. Experience brings a clear premium; engineers often double their salary after three to five years of work.

Embedded engineers have low replaceability because troubleshooting complex hardware systems and timing issues requires deep, hands‑on expertise that cannot be quickly substituted.

Choosing a Path: Long‑Term Viability Over Hype

The most scarce talent is engineers who can deliver complete products, integrating chips, circuits, and software, rather than those who only tune models in a notebook.

Embedded development provides tangible outcomes—code directly drives motors and LEDs—offering a sense of accomplishment that incremental loss reductions in algorithms do not match.

Career longevity is higher in embedded roles; the industry values accumulated experience, whereas algorithm positions often face replacement after age 35, especially if AI investment contracts.