AI Engineer

PERSOL Singapore; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

PERSOL Singapore is seeking an experienced AI engineer to support the production and continuous improvement of scalable AI/ML models, including multimodal LLMs. You will build ETL pipelines for large multimodal data, design evaluation benchmarks, and optimize deployment efficiency across cloud and edge platforms.

Mandarin fluency is essential for collaboration with China HQ. The role requires a Master’s in a technical field and at least 5 years of hands-on AI experience with PyTorch or

Qualifications

• Master's degree in CS/EE or related technical field.
• Minimum 5 years of AI engineer experience covering model training, evaluation and deployment.
• Strong foundation in statistics and deep learning theory.
• Proficiency in Python and at least one DL framework (PyTorch or MindSpore).
• Hands-on experience with multimodal model pre-training and large-scale data processing.
• Experience with model compression and efficient inference on cloud/edge platforms.
• Experience working with external customers and partners.
• Mandarin proficiency is essential for China HQ collaboration.

Responsibilities

• Support production and improvement of scalable AI/ML models including LLMs.
• Build ETL pipelines for large multimodal data to deploy AI solutions.
• Design evaluation benchmarks and run experiments to improve models.
• Optimize model efficiency via quantization, distillation, and pruning.
• Collaborate with data, training, inference, and product teams.
• Work with Ascend NPU clusters and deployment platforms, validate inference.

Skills

AI model training

Model evaluation

Python scripting

Statistical methods

Mandarin proficiency

Education

Master's degree in Computer Science or related field

Tools

PyTorch

MindSpore

TensorRT

Job description

Key Responsibilities

• Support the production and continual improvement of scalable and optimised AI/ML models, including multimodal large language models (LLMs), covering supervised fine-tuning (SFT), knowledge distillation, and large-scale distributed pre-training on 1000+-card GPU/NPU clusters.
• Build algorithms and automated pipelines for the extraction, transformation and loading (ETL) of large volumes of unstructured multimodal data (text, images, video and audio) - including data quality inspection, data format transformation and version management - to deploy AI solutions from theoretical data science models into production, including real-time on-device inference scenarios.
• Design multimodal evaluation benchmarks and automated evaluation pipelines; run experiments to test the performance of trained and deployed models; identify and resolve bugs and quality issues in models, data and evaluation pipelines, using evaluation results as feedback to drive the next iteration of training data and strategies.
• Optimise model efficiency for deployment through quantization (PTQ/QAT), distillation and pruning; analyse MoE model inference workloads to support system-level optimisation.
• Work in a team setting and apply knowledge of statistics, scripting (Python) and deep learning frameworks (PyTorch, MindSpore), collaborating with data, training, inference and product teams.
• Work with the hardware and software platforms on which models are trained and deployed: run training on large Ascend NPU clusters; on deployment platforms (e.g. TensorRT), set up inference validation environments, confirm model structure and operator compatibility, and benchmark quantized models for accuracy and speed, covering both cloud and on-device inference.
• To collaborate and communicate directly with Chinese speaking officers, stakeholders, counterparts, and customers in China HQ team

Required Skills & Qualifications

• Master's degree in Computer Science, Automation, Control Engineering, Electronic Engineering or a related technical field.
• Minimum 5 years of work experience in an AI engineer role covering AI model training, evaluation and deployment.
• Strong foundation in statistics, deep learning theory and model evaluation methodology.
• Proficiency in Python scripting and at least one mainstream deep learning framework (PyTorch or MindSpore).
• Hands-on experience with multimodal model pre-training, SFT and evaluation on large Ascend NPU clusters (1000+ cards), including evaluation benchmark design and large-scale unstructured data processing.
• Experience in model compression (quantization/distillation) and efficient inference deployment on cloud and edge platforms.
• Experience in direct technical engagement with external customers and technology partners.
• Proficiency in Mandarin is essential to ensure effective communication with the Chinese-speaking stakeholders, including head office, business counterparts, and customers based in China.

We regret to inform that only shortlisted candidates will be notified.