Advanced AI/ML Transformation Scientist

Mandate

Create materially different CPS workflows and operating models using advanced data and AI. Two profile levels are provided so Management and HR can match the vacancy to the depth of work and the local talent market. Both profiles require at least a master's degree; the advanced profile is intended for markets where research-calibre talent is realistically available.

Role purpose

Design, build and validate advanced AI-enabled workflows that can materially change how CPS services are delivered. This profile is intended for research-intensive transformation assignments requiring frontier modelling, rigorous experimentation and the ability to translate research into practical operating-model change.

Key responsibilities

• Work with end users and operational leaders to define transformation problems, current constraints and target outcomes.
• Create scalable and validated representations of structured, unconventional and unstructured data.
• Design, train, fine-tune and evaluate foundation, language, vision, multimodal and agentic models where justified by the use case.
• Develop new workflows in which AI performs defined work and qualified humans retain appropriate verification and decision authority.
• Back-test models and document accuracy, efficiency gains, failure modes, barriers to improvement and control requirements.
• Establish evaluation frameworks covering model performance, safety, guardrails, commercial impact and operational readiness.
• Translate complex methods, assumptions and limitations into recommendations that management and non-technical stakeholders can act upon.
• Work with regional and technology teams to assess reuse, implementation and scale across CPS.

Required qualifications and experience

• Master's, MPhil or PhD in Computer Science, Artificial Intelligence, Machine Learning, Statistics, Mathematics, Physics, Engineering or another highly quantitative field. Postdocs can apply as well.
• Advanced training and a strong applied research track record in machine learning, deep learning, NLP, computer vision, multimodal AI, generative AI or related methods.
• Strong hands-on programming capability in Python; C++ or another high-performance language is advantageous.
• Demonstrated experience designing reproducible experiments, evaluation frameworks and back-testing approaches.
• Ability to independently drive complex and open-ended work while collaborating with business, operations and technology teams.
• Strong English communication skills for an international working environment.

Preferred qualifications and experience

• First- or second-author publications at recognised venues such as ICML, ICLR, NeurIPS, KDD, AAAI, UAI, CVPR, ICCV, ACL, EMNLP, SIGMOD or equivalent.
• Experience with document AI, OCR, vision models, foundation-model fine-tuning, RAG, autonomous agents or Model Context Protocol.
• Experience with AWS Bedrock, Azure OpenAI, OpenSearch/Elasticsearch, Kafka, Redis or comparable enterprise platforms.
• Exposure to MLOps/LLMOps, model and prompt versioning, observability, guardrails and regression pipelines.
• Background in a regulated, quality-critical, laboratory, finance, hospital or TIC environment.

Indicative first-year outcomes

• A validated transformation methodology and target workflow linked to a defined business outcome.
• Reproducible evidence showing model performance, operational value, limitations and human-control requirements.
• A clear build, buy, partner, scale or stop recommendation accepted by sponsors.
• Reusable research assets, evaluation methods and documented knowledge transfer.

Hiring adaptation: This profile is best suited to Hong Kong, China or other markets where management expects to access research-calibre candidates. Publications should be preferred evidence of research quality rather than the sole selection criterion.