Physical AI (Edge Intelligence) Research Scientist

Physical AI (Edge Intelligence) Research Scientist About the Role We are seeking a Physical AI (Edge Intelligence) Research Scientist to advance research at the intersection of Physical AI, Embodied AI, Multimodal AI, Foundation Models, Robotics, and Edge Intelligence . The role focuses on developing AI models that can perceive, reason, and generate actions in real-world physical environments, with an emphasis on efficient, robust, and low-latency AI inference on edge systems . You will bridge the gap between AI research and real-world deployment , working closely with AI researchers and hardware acceleration engineers to develop novel algorithms, optimise AI models, and validate them on physical edge platforms and testbeds. Key Responsibilities • Conduct research in Physical AI, Embodied AI, Multimodal AI, and Foundation Models for real-world applications. • Develop and fine-tune Vision-Language Models (VLMs), Vision-Language-Action (VLA) models, multimodal models, and action-generation policies . • Develop AI algorithms for perception, reasoning, decision-making, and action generation using multimodal sensory inputs. • Optimise AI models for efficient and low-latency edge inference , including model compression, quantisation, pruning, knowledge distillation, and computational graph optimisation. • Deploy and benchmark AI models on edge devices, embedded platforms, robots, and physical testbeds . • Integrate vision feeds and multimodal sensor inputs into real-time AI inference pipelines. • Utilise simulation environments for synthetic data generation, pre-training, domain adaptation, and sim-to-real evaluation . • Collaborate with hardware and systems engineers to optimise AI workloads for target edge platforms. • Develop and optimise AI inference pipelines using PyTorch, ONNX, TensorRT , and related technologies. • Conduct experiments, benchmark model performance, analyse results, and translate research concepts into working prototypes. • Publish research findings in leading AI/ML, computer vision, robotics, and related conferences or journals. • Identify opportunities for technical patents and intellectual property . Qualifications • Ph.D. in Computer Science, Artificial Intelligence, Computer Vision, Robotics, Electrical Engineering , or a related quantitative discipline. Candidates with a research-oriented Master's degree and strong publications may also be considered. • Strong research background in one or more of the following areas: • Physical AI / Embodied AI • Robotics / Robot Learning • Multimodal AI • Vision-Language Models (VLMs) • Vision-Language-Action (VLA) Models • Foundation Models • Computer Vision • Edge AI • Strong understanding of Deep Learning, Transformers, multimodal models, or generative AI . • Strong publication record in top-tier AI/ML, computer vision, robotics, or related conferences and journals. • Strong proficiency in Python and PyTorch . • Familiarity with model deployment and optimisation frameworks such as ONNX, TensorRT, CUDA , or similar technologies. • Experience with model compression, quantisation, pruning, knowledge distillation, or inference optimisation is highly desirable. • Experience deploying AI models on edge devices, embedded systems, robots, or physical testbeds is highly desirable. • Working knowledge of C/C++ is preferred. • Strong research, experimentation, analytical, and scientific communication skills. Preferred Technical Experience Experience in any of the following areas would be highly valued: Physical AI | Embodied AI | VLA | VLM | Multimodal AI | Robotics | Robot Learning | Computer Vision | Edge AI | AI Inference | Model Optimisation | Quantisation | TensorRT | ONNX | Sim-to-Real | Sensor Fusion | Real-Time AI Research Focus The role covers the following research pipeline: Foundation Models → Multimodal Perception → Reasoning & Decision Making → Action Generation → Model Optimisation → Edge Deployment → Physical Testbed Validation The position is primarily focused on AI research, algorithm development, model optimisation, and edge deployment , rather than traditional telecommunications or semiconductor IC design.