Research Scientist I (CFAR), IHPC

CFAR, A*STAR is seeking an innovative Research Scientist to contribute to the development of cutting-edge AI technologies in preference learning, value alignment, and resource-efficient large language model (LLM) adaptation. This role focuses on advancing methods for fine-grained alignment of LLMs to heterogeneous societal preferences, particularly in multicultural environments like Singapore. The candidate will work on data collection, algorithm development, and resource-efficient deployment strategies to enable personalization of LLMs on local devices.

Successful candidates will be responsible, but not limited to:

Develop novel preference learning models that account for heterogeneous populations, ensuring fairness and inclusivity in LLM alignment.
Design and implement resource-efficient alignment techniques, including inference-time methods and parameter-efficient fine-tuning.
Construct and analyze Singapore-centric alignment datasets, capturing cultural, racial, religious, and linguistic diversity.
Develop evaluation benchmarks to assess LLM alignment quality across diverse sub-groups.
Collaborate with interdisciplinary teams to integrate domain knowledge from linguistics, social sciences, and computational AI.
Publish research findings in top-tier conferences and journals.

ABOUT A*STAR AND CFAR

At A*STAR, we make it our mission to attract and develop a diversity of talent along the research, innovation, and enterprise value chain, with career paths developed for scientists, engineers, and entrepreneurs. In return, we commit to investing in the personal and professional growth of each of our officers.

CFAR has a team of exemplary AI researchers committed to solve most exciting problems in the future AI, including IEEE fellows, chairs, scholars, and advisors among some of the most highly cited researchers in the world. Our AI Centre provides a unique research environment for fundamental AI research and cultivate next-gen AI scientists. If you are passionate about advancing AI alignment in multicultural societies and enabling scalable, efficient LLM adaptation, we encourage you to apply.

JOB REQUIREMENT

PhD in Computer Science, Computational Mathematics, Machine Learning, or related fields.
Strong background in AI, deep learning, and preference learning, with expertise in areas such as unsupervised learning, generative models, or controlled text generation.
Solid foundation in mathematics, capable of independent theoretical proofs, algorithm design, and integration with machine learning, optimization, and statistical modeling.
Experience with LLM alignment techniques, such as Reinforcement Learning with Human Feedback (RLHF) or Direct Preference Optimization (DPO).
Proficiency in developing efficient AI models, including parameter-efficient fine-tuning or inference-time adaptation methods.
Strong publication records in prestigious conferences and journals, such as NeurIPS, ICML, ICLR, AAAI, IJCAI, JMLR, TPAMI, AI, TNNLS, TKDE, TCYB.
Ability to work independently and collaboratively within interdisciplinary research teams.
The above eligibility criteria are not exhaustive. A*STAR may include additional selection criteria based on its prevailing recruitment policies. These policies may be amended from time to time without notice. We regret that only shortlisted candidates will be notified.
Type of Employment : Full-Time
Minimum Experience : 1 Year
Work Location : Immunos