Kerja Sepenuh Masa, AI Solutions Engineer di Superb AI Inc. - Maukerja

The Opportunity

Singapore is where the world's most demanding industrial operators validate AI. A reference deployment here is credibility everywhere. That is why Superb AI is building its APAC hub in Singapore.

We are already in production here: a vision AI safety system at a shipyard on Jurong Island, contracted work in the defence sector, and active engagements with port, airport, and advanced manufacturing operators. Over the next three years, Singapore grows into a regional hub building world-class applied AI solutions for these industries.

You would be the first engineer in that plan: the person who makes our technology work on customer sites in Singapore and across the region, with the authority to say what is feasible and what is not.

About the Job

You own the technical side of every customer and partner engagement in Singapore, from proving feasibility during the sales phase to fine-tuning and supporting what we deploy.

This is a customer-facing engineering role, and the work goes well beyond running demos.The sales lead is in the room with you, HQ engineering is one hour away, and the Global Lead reviews results before they go out. On most days, though, the technical judgement is yours.

The work splits roughly in half between pre-sales and delivery, built on Superb AI's own stack: ZERO, our vision foundation model; an MLOps platform for data curation, labeling, and retraining; and a video analytics layer. Much of this work now runs through agentic workflows and coding agents, and we expect you to work that way too.

Responsibilities

• Prove what works before we sell it. Run feasibility studies on customer data: build test sets that reflect real site conditions, report precision and recall per class, and write it up in a form the customer will actually read. Scope the data, accuracy targets, hardware, and timeline a deal commits us to. Lead technical workshops with customers and SI partners.

• Make deployed systems better every month. Fine-tune and deploy models on live projects. Run improvement cycles: collect field data and operator feedback, benchmark, retrain, verify, roll out, report. Own first-line support after go-live, and hand HQ a reproducible case when something needs escalation.

• Equip partners to deliver without us. Build reference implementations and train local SI partners on our tools so they can customise and deliver on their own.

• Send the field back to the product. Turn what you learn on site into feedback HQ product and research teams can act on.

• First 90 days: shadow an active engagement, run one feasibility study end to end with a review before results go out, take over first-line technical support for the region, and lead your first customer workshop with the sales lead beside you.

Minimum Qualifications

• At least 3 years of hands-on experience as a computer vision or ML engineer, or as a solutions engineer with real ML depth, with time spent in front of customers
• Fine-tuned detection or segmentation models on real-world data
• A habit of honest evaluation: you build test sets that reflect site conditions, report per-class precision and recall, and say "this doesn't work yet" when it doesn't
• Comfort debugging on someone else's infrastructure: Linux, Dockerm, or GPU servers
• Strong writing. Your feasibility write-up is often what the customer actually reads
• Composure in front of a customer when a number comes in below target
• Professional fluency in English to communicate with global team members and customers

Preferred Qualifications

• At least one model you took to production and kept improving
• Working knowledge of the NVIDIA stack: TensorRT, DeepStream, Jetson, or VSS. You know where each piece fits; expertise is not required
• Experience building agentic workflows (LLM or VLM pipelines that use tools) and daily use of coding agents
• Experience in defence, maritime, logistics, or industrial safety settings

About Superb AI

Superb AI builds vision AI for industrial operations, from the vision foundation model down to the platform its training loops run on.

Founded in 2018 and headquartered in Seoul with offices in San Mateo and Tokyo, we work with industrial leaders including Samsung, Hyundai Motor Group, Hanwha, Toyota, Nippon Steel, and Sharp, alongside NVIDIA and AWS. Our stack: ZERO, a multimodal vision foundation model pre-trained on industrial environments; an MLOps platform for data curation, labeling, and retraining; and a video analytics layer for safety monitoring and quality inspection. NVIDIA named Superb AI a Physical AI ecosystem partner at CES and GTC 2026.

Most vision AI companies deliver a model. We deliver a system that keeps improving after it ships. That is the work you would be doing in Singapore.

How we work

You are part of a global solutions function that uses the same method in every region. Account priorities and customer commitments sit with the regional sales lead. The quality bar and what goes to a customer sit with the Global Lead, who reviews feasibility results and benchmarks before they leave the building. Model architecture changes, platform bugs, and anything beyond a field fix go to HQ engineering in Seoul.

Where this role goes

Year one is about earning trust: feasibility studies the customer believes without HQ in the loop, at least one improvement cycle on a live system with the numbers reported straight, and one SI partner customising on its own.

Once the first deployments are stable, the scope widens. Singapore's flagship industries need solutions that do not yet exist as products. Using ZERO and our agentic stack, you will start shaping those solutions for the local market, directly or with partners. The ones that work become references we take across APAC.

As the Singapore team grows, so does this role. The engineer who built the first deployments is well-placed to shape what the regional technical team becomes.