Home >
HR Insights >
Physical AI Hiring in Singapore: The Talent Bottleneck Is Moving to the Hardware-Software Boundary

Physical AI Hiring in Singapore: The Talent Bottleneck Is Moving to the Hardware-Software Boundary

October 5, 2026
HR Insights

Share this article:

For the last few years, much of the conversation around artificial intelligence has focused on what happens on a screen.

Chatbots. Copilots. Generative AI. AI agents.

The next wave is starting to acquire a physical form.

Physical AI, sometimes described as embodied AI, brings intelligence into machines that can sense their environment, make decisions and act in the real world. That includes autonomous mobile robots, humanoids, robotic arms, drones, autonomous forklifts and increasingly intelligent industrial systems.

Singapore is taking this seriously.

In May 2026, IMDA, JTC and the Singapore Institute of Technology announced plans for a real-world physical AI testbed at Punggol Digital District, working with eight industry leaders. The environment is intended to support multiple robotics use cases and operators at scale, including applications in delivery, cleaning and security.

The National Robotics Programme is Singapore’s national initiative for robotics and embodied AI development and innovation.

But as physical AI moves from research to deployment, another problem is becoming increasingly visible.

Who is going to build, integrate, debug and maintain all of it?

Physical AI creates a different kind of talent problem

Hiring for physical AI is not simply a matter of finding more “AI engineers”.

The systems being developed need intelligence, but they also need motors to move, sensors to perceive, circuit boards to function, embedded software to respond on time, mechanical systems to survive the physical world and infrastructure to connect everything together.

That creates demand across disciplines such as:

Engineering & hardware

Robotics engineering
Electrical and electronics engineering
Mechanical engineering
Mechatronics
PCB and PCBA design
Sensor integration
Industrial automation

Software, autonomy & infrastructure

Firmware and embedded systems
Computer vision and perception
Navigation and controls
Autonomous systems
Edge computing
Real-time systems
Cloud and distributed infrastructure

The interesting part is what happens between these disciplines.

A mechanical engineer working on a robot may need to understand sensor placement and electrical constraints.

An electrical engineer may need to understand firmware behaviour and control systems.

A firmware engineer may need to read schematics, debug hardware signals and understand what the mechanical and controls teams are trying to achieve.

A robotics software engineer may need to understand what happens when algorithms leave simulation and encounter imperfect sensors, changing loads and unpredictable physical environments.

Physical AI therefore creates demand not only for specialists, but for engineers who can operate across boundaries.

What Tyson Jay is seeing in live hiring

Our current recruitment work provides an early view of this shift.

As at 29 September 2026, Tyson Jay is supporting six active, published roles across two Singapore-based robotics and automation hiring programmes.

Five of those six roles, or 83.3%, are explicitly in robotics, hardware, firmware, electrical or mechanical engineering.

Only one is a conventional software engineering role.

That does not mean software is becoming less important. Quite the opposite. It suggests that as AI moves into physical systems, software increasingly depends on engineering disciplines around it.

September 2026 Hiring Snapshot

278

Candidate-job matches across six live hiring pipelines

Where the matches came from

Direct Applicants (68.7%)

Active sourcing (31.3%)

Candidate location

Recorded as Singapore-based (60.1%)

Point-in-time pipeline progression

4.7%

Reached client submission

2.5%

Reached interview

Source: Tyson Jay recruitment pipeline data as at 29 September 2026. Figures represent candidate-job matches rather than unique individuals. Searches remain active, so submission and interview figures are point-in-time pipeline measures, not completed hiring conversion rates.

Those last two figures are not final conversion rates. The searches remain active and a large proportion of the pipeline is still awaiting assessment. But the early pattern is important.

The challenge is not necessarily attracting people. It is finding the right combination of capabilities.

Having hundreds of candidate records does not automatically create a deep pool of engineers who can design, integrate, validate and troubleshoot complex physical AI systems.

This is consistent with the wider Singapore market

Tyson Jay’s dataset is small and should not be treated as representative of the entire Singapore labour market.

But the direction is consistent with much larger datasets.

According to IMDA’s Singapore Digital Economy Report 2025, the share of Singapore tech job postings requiring AI skills increased from 11% in 2019 to 14% in 2024.

Manufacturing is particularly relevant to physical AI.

The number of manufacturing tech job postings requiring AI-related skills increased from approximately 390 in 2019 to 840 in 2024, more than doubling over five years.

Singapore’s Ministry of Manpower has separately reported that engineering and technology-related occupations remain among the roles with strong labour demand.

This matters because physical AI sits directly at the intersection of these areas.

It is AI meeting engineering.

Singapore is already preparing the next generation

The skills pipeline is also starting to respond.

A physical AI training initiative supported by the National Robotics Programme announced in 2026 aims to expose at least 10,000 students to physical AI over five years, ranging from primary school students to university students.

The focus goes beyond learning to code.

Students are expected to work on robotics, autonomous systems and real-world problem solving.

That distinction is important.

The physical AI economy will certainly need machine-learning researchers.

But it will also need people who know how to make an autonomous machine work reliably at 3pm in a warehouse when a sensor fails, a communications link becomes unstable or the physical environment behaves differently from the simulation.

Those skills are developed through engineering experience, not merely through familiarity with an AI model.

The jobs may change before the job titles do

There is another trend employers should watch.

Not every physical AI job will have “AI” in its title.

A Firmware Engineer can be part of an AI robotics team.

A Hardware Engineer can determine whether an autonomous system survives real-world deployment.

A Mechanical Engineer can influence perception performance through sensor placement.

A Solutions Engineer may be responsible for translating robotics capabilities into a working customer deployment.

An AGV technician may increasingly work with autonomous systems, sensors, software diagnostics and fleet-management platforms.

This is similar to what Singapore is already experiencing with generative and agentic AI.

Jobs can change substantially even when their titles remain familiar.

For employers, searching only for candidates who already have the exact future job title could therefore make an already constrained talent pool even smaller.

The strongest candidates may come from adjacent industries

One of the opportunities we see in recruitment is to look beyond obvious labels.

A candidate may never have worked on a humanoid robot but could have highly transferable experience from:

Industrial systems & mobility

Industrial automation
Autonomous vehicles
Semiconductor equipment
Industrial machinery
Warehouse automation
Motion control

High-reliability engineering

Medical devices
Aerospace
Defence engineering
Automotive electronics
Embedded systems
High-reliability electronics

Someone who has designed and debugged embedded control hardware for industrial machinery may be closer to a robotics electrical engineering requirement than someone whose job title happens to contain the word “robotics”.

This is where skills-based hiring becomes particularly important.

Employers may need to rethink the “perfect candidate”

Physical AI also presents a familiar hiring problem in a more extreme form.

Employers naturally want someone who has already done everything.

A single job description can easily accumulate requirements across:

electronics, PCB design, embedded systems, robotics, mechanical integration, controls, Python, C++, Linux, CAN, ROS, sensors, manufacturing and field deployment.

The resulting candidate may exist.

But there may be very few of them.

And every physical AI company is competing for the same people.

A December 2025 CNA report noted that some robotics positions can remain vacant for more than six months.

The alternative is not to lower standards.

It is to identify which capabilities are genuinely foundational and which can be learned.

For example, an employer could distinguish between:

Must already know

Real-time embedded systems and hardware debugging

Can learn

The company’s specific robotic platform.

That distinction can significantly increase the addressable talent pool without compromising engineering quality.

Hiring speed is going to matter

Physical AI talent is not being pursued only by startups.

Singapore is attracting companies working across robotics, AI, advanced manufacturing and autonomous systems.

Sharpa, for example, announced plans in 2026 to expand its Singapore operations and recruit dozens of AI scientists, mechatronics engineers and solutions engineers.

At the same time, physical automation is spreading into established sectors.

HDB is scaling robotics and automation across construction sites. Logistics operators are introducing autonomous warehouse systems. Robots are increasingly being deployed in cleaning, security, delivery and industrial environments.

That means the same engineering capabilities can be relevant to many industries.

Companies that take six weeks to decide whether to conduct a first interview may increasingly discover that the candidate has already moved.

Physical AI may create an experience bottleneck

Singapore’s challenge may therefore not simply be a shortage of graduates.

It may be a shortage of people with enough real-world engineering repetitions.

Someone must have debugged the board that did not boot.

Someone must have discovered that the sensor position worked in CAD but failed in the real environment.

Someone must have tuned the system when timing, temperature, vibration or network latency behaved differently from expectations.

Someone must have brought a prototype through validation and deployment.

Those experiences take time to accumulate.

Physical AI could therefore intensify an existing problem in technology hiring: companies compete aggressively for experienced engineers while giving fewer junior engineers opportunities to acquire the experience everyone subsequently demands.

Singapore’s decision to expose thousands of students to physical AI is a useful beginning.

Companies will need to complete the pipeline by creating meaningful opportunities for those engineers to work on real systems.

The opportunity for Singapore

Singapore is unlikely to compete with much larger countries on engineering headcount alone.

It does not need to.

Its opportunity may lie in becoming one of the best places in the world to bring advanced physical AI systems from research into reliable real-world deployment.

That requires more than AI researchers.

It requires an ecosystem of electrical engineers, mechanical engineers, firmware developers, robotics engineers, technicians, systems engineers, AI researchers and operators who understand how to make intelligent machines work safely and reliably around people.

The technology conversation around physical AI is accelerating.

The hiring conversation needs to catch up.

Because the next AI talent shortage may not be about who can build the model.

It may be about who can make the model work in the physical world.

Frequently Asked Questions

What is Physical AI?

Physical AI, sometimes described as embodied AI, refers to intelligent systems that can sense their environment, make decisions and act in the physical world. Examples include autonomous mobile robots, robotic arms, drones, humanoids and intelligent industrial systems.

What roles are needed to build Physical AI systems?

Physical AI teams can require expertise across robotics, electrical and electronics engineering, mechanical engineering, mechatronics, firmware and embedded systems, computer vision, controls, autonomous systems, edge computing and infrastructure.

Does Physical AI hiring only require AI or machine-learning engineers?

No. AI may provide the intelligence behind a system, but successful deployment also depends on hardware, sensors, embedded software, mechanical systems, controls and infrastructure. Many of the most important roles may not have ‘AI’ in their job title.

How can employers widen the talent pool for robotics and Physical AI roles?

Employers can look beyond exact job titles and assess transferable engineering experience from areas such as industrial automation, semiconductor equipment, autonomous vehicles, medical devices, aerospace, embedded systems and other high-reliability engineering environments.

Why can Physical AI roles be difficult to fill in Singapore?

The challenge is often not simply finding engineers, but finding candidates with the right combination of cross-disciplinary and real-world deployment experience. Some roles require expertise across hardware, firmware, controls, sensors and software, which can significantly narrow the available talent pool.

About the data

Tyson Jay analysed six active, published recruitment assignments related to robotics and automation across two Singapore-based employers as at 29 September 2026. The dataset contained 278 candidate-job matches. A candidate can be matched to more than one role, so these figures should not be interpreted as 278 unique individuals. Recruitment activity remains ongoing.