AI算法工程师（HK）

Salary: $25,000 – $30,000 per month

About the role

You will be responsible for AI model work spanning data preparation, training and fine-tuning, effect evaluation, and deployment on cloud or edge devices (depending on your expertise and interests). You will collaborate with product, algorithm, software, hardware, and testing teams to transform research outcomes into stable and usable product capabilities.

Key responsibilities

• Develop technical roadmaps for AI models, conduct model selection and effect optimization, perform model training, fine-tuning, inference and validation, and choose appropriate implementation solutions based on product requirements.

• Build data and training pipelines, participate in establishing data formats, metadata, annotation and version standards, and drive collection, generation, quality assurance, procurement and compliant use of multi-source data.

• Establish unified model evaluation and regression systems, compare third-party, open-source and proprietary models using real device data, and analyze task quality, robustness, latency, stability, resource consumption and usage costs.

• Drive model deployment on cloud and edge devices, complete model conversion, quantization, compression, pruning, distillation and inference acceleration, and optimize solutions based on target device constraints in computing power, memory, power consumption and latency.

• Complete edge deployment using frameworks such as ONNX, TFLite, NCNN, MNN, or perform online inference, auto-scaling, version release and rollback based on GPU, containers and cloud services.

• Collaborate with signal processing and systems teams to optimize end-to-end performance, analyze the impact of data collection, pre/post-processing and runtime environment on model performance, and assist in locating end-to-end issues.

• Use internal AI engineering platforms to manage datasets, training tasks, model versions, validation reports, deployment and runtime logs, ensuring experiments are reproducible and model versions are traceable, releasable and rollbackable.

• Work with product, engineering and testing teams to complete technical research, solution design, feature integration and production acceptance, and address practical issues between experimental results and user experience.

About you

• Bachelor's degree or above in Computer Science and Technology, Artificial Intelligence, Electronic Information, Communication Engineering, Signal Processing, Acoustics, Automation or related fields.

• Relevant experience in at least one of: model training, multimodal understanding, generative AI, intelligent interaction, signal processing, or edge inference.

• Familiar with training tools such as PyTorch and Transformers, able to participate in data processing, model training, fine-tuning, inference, error analysis and experiment reproduction.

• Proficient in Python with C or C++ development capabilities. Prior experience with Android, iOS, embedded Linux, DSP or mobile model deployment is preferred.

• Understand metrics such as task quality, accuracy, recall, real-time rate and end-to-end latency, able to design controlled experiments and identify actionable improvements from error cases.

• Prior experience with model quantization, pruning, distillation, compilation optimization, GPU inference optimization, or practical projects using ONNX, TFLite, NCNN, MNN is preferred.

• Prior experience with Azure Machine Learning, MLflow, Docker, Kubernetes, model serving or MLOps platforms is preferred.

• Prior experience with audio signal processing, multi-channel perception, embedded algorithms or resource-constrained device optimization is preferred.

• Able to transform research conclusions into code, test results and product solutions, willing to engage deeply with real devices, real audio and complex usage scenarios.

• Maintain curiosity about new technologies, value experimental evidence, engineering quality and team collaboration, able to clearly articulate solution effectiveness, limitations and risks.