10 model-based-control Fellowship positions at SINGAPORE INSTITUTE OF TECHNOLOGY (SIT)
-
(Lua/Java), agent behavior modeling, event handling, and API-based integration with external AI systems. Experience with distributed systems, reinforcement learning, or simulation environments (e.g
-
pipelines, cloud services, dashboards, web-based application development). Experience in handling real-world datasets and model development/deployment workflows. Domain expertise in building construction
-
Model Development Support development and evaluation of learning-based control policies using reinforcement learning, imitation learning, and visuomotor learning approaches. Assist in model training
-
. Familiarity with immersive platform development (e.g., Roblox, Minecraft), including scripting (Lua/Java), agent behavior modeling, event handling, and API-based integration with external AI systems. Experience
-
software engineering standards through rigorous testing, version control, clear documentation, and continuous improvement of development infrastructure (e.g., CI/CD, Docker, cloud-based services). Carry out
-
. Implement domain randomization and scenario variation pipelines for robust sim-to-real learning. v. AI Model Development Support development and evaluation of learning-based control policies using
-
fine-tuned Vision-Language Models (VLMs) for game scenario detection, supporting zero-shot reasoning and scene-graph inference. Ensure the system is deployment-ready by supporting benchmarking
-
AI models to automate QA/QC processes, enabling real-time verification, anomaly detection and consistency checks across records. Develop a cloud-based verification platform for digital data collection
-
testing, version control, clear documentation, and continuous improvement of development infrastructure (e.g., CI/CD, Docker, cloud-based services). Carry out Risk Assessment, and ensure compliance with
-
narratives) Leverage fine-tuned Vision-Language Models (VLMs) for game scenario detection, supporting zero-shot reasoning and scene-graph inference. Ensure the system is deployment-ready by supporting