Technical search for applied AI in Vietnam.
Applied AI in Vietnam is being built inside foreign-invested plants and the engineering centres attached to them. Samsung's research centre in Hanoi is the largest the group runs in Southeast Asia. LG operates display and component plants in Hai Phong with engineering teams beside them. Intel's assembly and test campus in Ho Chi Minh City, Amkor in Bac Ninh, and the EMS and ODM plants run by Foxconn, Luxshare, Goertek and Pegatron across the northern provinces all carry the same pressure: more inspection, tighter traceability, and more data coming off the line than people can read. That pressure is what turns AI from a corporate initiative into an engineering job with a deadline attached.
The work is applied in the literal sense. A model inspects a solder joint, guides a robot, predicts when a machine will fail, or is verified as part of an accelerator before tape-out. Chip design centres run by Marvell, Synopsys, Renesas and Ampere brought that last kind of work into the country, and Bosch's software operation in Ho Chi Minh City brought the automotive version of it. What makes these roles different from AI work elsewhere is where they start. Most plants here have no clean dataset waiting to be used. The engineer instruments the line, collects the images, labels them, and only then begins the part most people would call the job. The result is a generation of engineers who understand the machine as well as they understand the model.
This is a young market and it is moving quickly. NVIDIA's agreement with the government to establish an AI research and development centre, FPT's AI factory, and the data centre capacity now being built give the country something it did not have three years ago: somewhere to train models, not only to deploy them. The national AI strategy and the semiconductor training programme are pointed at the same decade. The people who will lead these teams in ten years are mid-level engineers today, and most of them are inside an FDI plant or a design centre right now. Kelson searches in that layer.
Roles we search for
AI silicon and accelerators
AI hardware architect · NPU and AI accelerator design engineer · AI accelerator verification engineer · Performance modelling and analysis engineer · Hardware-software co-design engineer
Edge and embedded AI
Embedded AI engineer · Edge AI engineer · Model optimisation and quantisation engineer · AI firmware engineer · Embedded Linux and BSP engineer · DSP and signal processing engineer
AI compilers and runtimes
Compiler and toolchain engineer · GPU kernel and library engineer · Inference runtime engineer · ML framework engineer
Computer vision and inspection
Computer vision engineer · Deep learning engineer · Image processing engineer · Vision algorithm developer · AI inspection systems engineer · Defect classification specialist · Vision systems integration engineer
Machine learning and industrial data science
Machine learning engineer · Data scientist · Applied research engineer · Predictive maintenance engineer · Anomaly detection engineer · Sensor and time series analytics engineer · Process optimisation modeller · Operations research and optimisation engineer
AI infrastructure and platforms
MLOps engineer · ML platform engineer · AI infrastructure engineer · GPU cluster engineer · HPC engineer · Distributed training engineer · Model serving engineer · Edge computing engineer
Data engineering
Data engineer · Data platform architect · Industrial data pipeline engineer · Analytics engineer · Data annotation lead · Data governance lead
AI deployment and solutions
AI deployment engineer · AI solutions architect · AI application engineer · Technical programme manager · Customer solutions engineer
AI in operations
Manufacturing data science manager · Supply chain forecasting lead · Demand planning lead · Quality analytics manager · Business intelligence lead · Digital transformation manager
Leadership
Head of AI · Head of AI engineering · Head of data · AI infrastructure director · R&D director for AI · Principal AI scientist · AI transformation lead · AI programme director · Chief data officer
These searches start at a gross monthly package of around 50 million VND, or below that when several roles are open at once.
What we look at
Whether the system is still in use, and how long it has been running. The numbers after six months of real production, not the ones from the trial. Where the training data came from, and who built and labelled it. For inspection: the false call and escape rate once the line was running at volume. For embedded and edge roles: what the model had to run on, and what was traded away to make it fit. For compiler and runtime roles: which stack they worked in, TensorRT, ONNX Runtime, OpenVINO or TVM, and whether they wrote kernels or called them. For machine learning and industrial data science: what the model predicted, how far ahead, and whether anyone changed a decision because of it. For infrastructure: what runs unattended, and at what scale. For deployment roles: how many sites they rolled the same system out to, and what broke on the second one. For accelerator work: which block they owned, and how the silicon behaved when it came back.
Most of the AI market in Vietnam sits in software services, fintech and consumer products — a large, fast-moving market Kelson does not work in. Inside what is left, the overlap between people who understand models and people who understand the physical thing a model runs on is small, and more often it is a hardware or manufacturing engineer who learned ML than a data scientist who learned the plant. The pool is still small, though it is deepening faster here than in most markets. Senior roles often mean looking at Vietnamese engineers abroad who are ready to come back or international engineers, and we map that out with you before the search begins.