AI Safety Research Associate (Computational Linguistics/Computer Science)

Nanyang Technological University; Região Norte; Presencial; BRL 326 000 - 530 000; Tempo integral

Resumo da oferta

NTU Singapore is seeking a research-oriented candidate to advance safety and alignment for LLMs in the medical domain. You will co-design guardrails, analyze failures in SEA languages, and contribute to regional language capabilities within AISG-hosted initiatives at NTU.

The role involves building auditable reporting dashboards and co-authoring publications for top conferences in AI safety and NLP. Strong collaboration and Python/NLP skills are essential.

Qualificações

• Master degree or equivalent practical experience in computational linguistics, CS or related field.
• Experience in safety alignment for LLMs and model evaluation for safety or robustness.
• Track record of publications in safety research is required.
• Strong Python and NLP fundamentals are essential.
• Hands-on with LLM inference frameworks and generative AI tools is required.

Responsabilidades

• Co-design safety and alignment strategies for LLMs in the medical domain.
• Co-develop guardrails and safeguard models.
• Analyze failure modes for SEA languages and cultural contexts.
• Support regional language capabilities to extend SEA-LION usefulness.
• Identify culturally grounded safety risks and ensure robust generalisation for SEA languages and code-mixed text.
• Design leaderboards and visualisations for auditable reporting of results.
• Co-author top-tier publications to contribute auditable, reproducible findings.

Conhecimentos

Python

NLP concepts

LLM inference frameworks

Generative AI

Leaderboards & visualisations

Communication

Formação académica

Master Degree in Computational Linguistics or Computer Science

Ferramentas

Ollama

vLLM

Descrição da oferta de emprego

AI Singapore (AISG) is a national AI programme launched by the National Research Foundation (NRF), Singapore, to build and anchor deep national capabilities in AI. AISG is supported through a government-wide partnership including the NRF, Ministry of Digital Development and Information (MDDI), Infocomm Media Development Authority (IMDA), Economic Development Board (EDB) and Enterprise Singapore (ESG). We bring together research institutions and the vibrant ecosystem of AI start-ups and companies to support impactful research, develop talent, and power Singapore's AI efforts.

This position will be hosted at the Nanyang Technological University (NTU) under VP (Artificial Intelligence & Digital Economy)’s office and we welcome you to join our community.

Key Responsibilities:

• Co-designing and co-implementing safety and alignment strategies for LLMs in the medical domain

• Co-developing guardrails and safeguard models

• Assist in analysing failure modes specific to Southeast Asian languages and cultural contexts

• Supporting development of regional language capabilities to extend utility of SEA-LION for the region

• Assist in identifying culturally grounded safety risks and ensure our safety systems generalise robustly beyond English-centric assumptions for SEA languages and code-mixed text

• Assist in designing leaderboards and relevant visualisations for auditable reporting of results

• Co-authoring top-tier publications (e.g., ACL, EMNLP, NeurIPS, ICLR, etc.) to contribute novel, auditable, and reproducible findings

Job Requirements:

• Preferably a Master Degree in Computational Linguistics, Computer Science, or have equivalent practical experience.

• Required experience includes safety alignment in LLMs, model evaluation for safety or robustness, and a track record of publications in the safety research area.

• Strong knowledge of Python and NLP concepts is a must.

• Experience with LLM Inference frameworks (i.e., Ollama or vLLM) is a must.

• Experience with Generative AI and LLMs (i.e., Llama, Gemma, Qwen, etc) is a must, along with competence in designing leaderboards and visualisations.

• Good communication skills for our collaborative environment.

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