Senior Localization Engineer / Research Engineer, Robotics AI (Visual Navigation)

About the role

Grab's Robotics Technology team builds urban embodied AI, owning the product lifecycle from perception and navigation research to deployment on its autonomous delivery fleet. Based in Singapore, this Research Engineer position on the Spatial Intelligence team advances visual navigation and Vision-Language Navigation (VLN) capabilities so robots can understand and navigate complex urban environments. The role reports to the Head of Engineering and works onsite at the One North office.

Key responsibilities

• Design and implement deep learning models for visual navigation, VLN and visual place recognition in real outdoor environments, including multimodal pipelines combining RGB, depth, semantic segmentation and language-conditioned signals.
• Research scene understanding for dynamic urban settings such as pedestrian-dense streets, repetitive facades and GNSS-degraded corridors, and explore end-to-end learning and modular architectures for navigation policy.
• Publish research and share findings internally and in open-source repositories.
• Integrate and adapt state-of-the-art VLN and visual navigation algorithms (e.g., R2R, REVERIE, EmbodiedBERT derivatives) onto real robot hardware, and build evaluation frameworks for last-mile delivery scenarios.
• Work with Perception and Planning/Control teams to embed visual navigation outputs in the autonomy stack.
• Run closed-loop dataset collection and annotation, and partner with hardware and systems teams on sensor selection, calibration and data quality.

Required profile

• PhD in Computer Science, Robotics, AI, Computer Vision or a closely related field.
• Strong publication record or demonstrable research contributions in visual navigation, VLN, embodied AI or robot perception.
• Ability to design, run and critically evaluate experiments, with strong written communication for research reports and publications.

Required skills

• Proficiency in at least two of: VLN benchmarks (R2R, REVERIE, SOON) and instruction-following models; image-based retrieval, learned descriptors or topological navigation; CNNs, Transformers and multimodal architectures (ViT, CLIP, LLaVA variants) for robotics; reinforcement learning for navigation and sim-to-real transfer.
• Python and PyTorch (or JAX/TensorFlow); experience with ROS/ROS 2.
• Nice to have: deployed navigation models on physical robots, LiDAR-visual fusion, large-scale vision-language models, and prior work with an autonomous vehicle or robotics company.

What we offer

• Term life insurance and comprehensive medical insurance.
• GrabFlex benefits package.
• Parental and birthday leave, plus volunteering leave.
• A confidential assistance programme for employees and their families.