Research Scientist (11-month contract or till 9 June 2027), SCIS, Singapore

Description

Responsibilities

Construct SAR‑specific instruction‑following datasets that incorporate Visual Chain‑of‑Thought (VCoT) reasoning to teach the model problem decomposition.
Develop and fine‑tune Vision‑Language‑Large‑Models (VLLM) to interpret user queries and utilize the Expert ModelToolbox for tasks like detection and segmentation.
Implement Reinforcement Learning from Human Feedback (RLHF) to align the Co‑Pilot's generative policy with human expert preferences for utility and trustworthiness.
Integrate the SAR foundational model (from WP1) into the LLM architecture using lightweight projection networks to enable understanding of SAR‑specific visual representations.
Develop an interactive web‑based demonstration environment for real‑time interaction with the SCOPE agent.
Other duties as assigned.

Qualifications

Minimum PhD degree in Computer Science, Information Technology, Information Systems or closely related disciplines; with particular focus on deep‑learning based AI tech research.
Minimum 5 years of relevant research experience (including experience during PhD) in deep learning, multimodal data processing (especially visual data processing), and domain generalization/adaptation.
Proficiency in programming software/languages such as PyTorch and TensorFlow will be positively considered/ is required.
Proficiency in Claude Code or other mainstream Video Coding tools.
Proven track record and professional experience in deep learning and computer vision.
Demonstrated experience and know‑how on interactive AI techniques in multi‑modality data.
Ability to break down complex problems and design effective solutions.
Innovating new approaches and thinking outside the box.
Applicants with research publications in machine learning and computer vision areas will be advantageous.
Ability to work cooperatively as part of a small, agile academic research team is essential.
Self‑motivated individual who can work independently and collaboratively with team members.

#J-18808-Ljbffr

Highlights

Safety Tips

Be careful with commission-based ’work-from-home’ positions that offer an unrealistically high income.