At Social AI Studio, we aim to develop and train the future generation of social AI scientists/designers. Email us your CV!

• Postdoctoral Research Fellow: Candidate should have a Ph.D. degree and a promising academic record. Topics considered particularly relevant include social media mining, computational social science, machine learning, deep learning, recommender systems, natural language generation. A strong background in other computational social science, data mining and machine learning topics will also be considered. 

• Research Engineer: Candidate should have a Bachelor’s and/or Master’s degree in Computer Science or closely-related disciplines. He or she should be competent in at least one programming language and have some experience in implementing data mining algorithms and machine learning models (school projects included). 

• Master/Ph.D. Student: Candidate should have a Bachelor’s and/or Master’s degree in Computer Science or a closely-related discipline with high academic standing. He or she should be competent in at least one programming language and demonstrate a strong passion for AI and/or computational social science research. We also offer scholarships to outstanding students enrolled in our Ph.D. programs!

• Research Intern: Candidate should be currently enrolled in a Bachelor’s and/or Master’s degree in related disciplines. He or she should have a keen interest to learn more about AI and/or computational social science research.

The following funded projects are hiring Postdoctoral Research Fellows, Research Engineers, and Research Interns:

1. Detecting and Monitoring Hate Speeches in Social Media - This proposal aims to design effective data analytical tools and machine learning algorithms to curb the spread of hate speech on social media. Specifically, we outline two main aims of this project: (i) Develop semi-supervised meta-learning algorithms to detect hate speech on multiple social media platforms. (ii) Designing multilingual analytical monitoring dashboards that monitor hateful social media content across Southeast Asian (SEA), interpreting content in native southeast Asian languages.

2. SafeLens: Hateful Video Moderation with Large Multimodal Models - This project aims to develop innovative AI tools for detecting and mitigating hateful video content in low-resource languages prevalent in Singapore and Southeast Asia. Building on prior research in text-based moderation, it focuses on (i) creating annotated datasets for hateful videos in languages like Bahasa Melayu and Tamil, (ii) designing low-resource video moderation models capable of identifying and localizing hate speech in multilingual video content, and (iii) integrating explainability modules that provide cultural-contextual explanations for flagged videos. SafeLens advances video content moderation in multilingual contexts, promoting empathy, solidarity, and digital well-being across the Southeast Asia region.

3. Next-Gen Competitive Intelligence: Automated Discovery of Outstanding Facts with Large Language Models -  This project aims to revolutionize competitive analysis by leveraging LLMs to automate the discovery and evaluation of "Outstanding Facts" (OFs)—insights that reveal an entity’s competitive edge. Unlike traditional methods that depend on predefined data schemas and require intensive manual input, this project introduces a dynamic Scope-Competitor exploration model, enabling adaptive scope and competitor discovery using both LLMs and external data. Additionally, it employs a dual-layered evaluation framework combining LLM-generated assessments and human feedback through RLHF to ensure high-quality OF identification. Finally, a prototype system will be developed and tested across real-world scenarios, including company, labor market, and product competition analysis, showcasing the model’s practicality and broad applicability in economic and social contexts.

4. Heterogeneous Preference Learning for Value-Aligned Large Language Models in Multicultural Societies: A Resource-Efficient Approach - This project aims to develop fine-grained preference and value alignment techniques for LLMs that can be deployed on mobile devices, focusing on the multicultural context of Singapore. Unlike existing methods, which assume a homogeneous population and require intensive computational resources, this project will (i) compile a Singapore-centric dataset capturing diverse sub-group preferences and values, (ii) design mixture preference models that incorporate explicit demographic features, and (iii) implement efficient alignment approaches such as conditional diffusion models and low-rank adapters for mobile deployment. Ultimately, this project will advance personalized and locally deployable LLMs that respect user preferences and societal diversity.

5. LLM Architecture-Language Nexus: A Unified Analysis Framework - This project aims to develop a unified framework for systematically analyzing and advancing LLM architectures for multilingual tasks, with a special focus on Southeast Asian (SEA) languages. By leveraging a range of state-of-the-art architectures, including Transformers, emerging models like Mamba, and their hybrids, the project seeks to enhance the efficiency, scalability, and inclusiveness of multilingual LLMs, particularly for underrepresented SEA languages. Through a phased approach, it will explore architectural trade-offs, optimize training strategies (e.g., data scheduling and cross-lingual transfer), and produce real-world deployable models. The project aspires to empower SEA communities, foster social cohesion, and contribute to global AI advancements for low-resource languages, while working closely with AI Singapore, industry partners, and international collaborators.