Lead Edge AI Engineer

Get to know the Team

The Data Science (Geo Vision) team at Grab focuses on improving the maps and building map-based intelligence such as localization routing travel time estimation and traffic forecasting. We use Computer Vision and conventional machine learning methods on a variety of signalsspecifically utilizing edge device footageto understand our locations and road networks.

Get to know the Role

We are looking for a Lead Data Scientist / Edge AI Engineer to lead edge development for our edge devices. A key focus will be Multi-Task & Action Recognition Development where the successful candidate will be responsible for developing and refining multi-task learning models and video action recognition systems for our edge devices.

You will work onsite and will report to the Head of Data Science based in the Cluj Office.

The Critical Tasks You Will Perform

• Multi-Task & Action Recognition Development: Develop and refine multi-task learning models (specifically Hydranet architecture) and video action recognition systems using PyTorch.
• Edge Deployment & Engineering: Deploy Computer Vision algorithms into embedded Android platforms utilizing the Qualcomm SNPE / QNN SDK to interact directly with the DSP.
• Resource Efficiency: Conduct rigorous performance analysis to reduce power consumption and manage thermal constraints. You will ensure model switching latency remains minimal to maintain recording integrity.
• System Stability: Implement safety mechanisms to ensure system stability during dynamic model graph reconfiguration.
• Collaboration: Collaborate with Firmware and Mobile teams to integrate signals for model decision-making.

Qualifications :

What Skills you will Need:

• Educational Background: Bachelors degree or higher in Computer Science Electrical Engineering or a related field.
• Experience: Minimum of 5 years of experience in computer vision and machine learning with a mandatory focus on Edge AI deployment and optimization.
• Mandatory Frameworks: Deep expertise in PyTorch. You must be proficient in Video Action Recognition and Learning network design.
• Edge Constraints: Deep understanding of resource-constrained environments (memory bandwidth thermal management and power consumption).
• Optimization: Knowledge of general model optimization techniques for edge devices (quantization pruning graph splitting).
• Communication: Proficiency in English (speaking and writing) with the ability to present technical data insights.

The Nice-to-Haves:

• Context-Aware DSP Optimization & Dynamic Graph Execution: Experience designing and implementing intelligent runtime management logic (dynamic model switching or conditional computation) on Android to dynamically load/unload specific heads of multi-task architectures (e.g. Hydranet) on the Qualcomm Hexagon DSP.
• DSP & SNPE Expertise: Proven experience with Qualcomm DSP (Hexagon) SNPE (Snapdragon Neural Processing Engine) or QNN SDK.
• Android Development: Background in Android development for edge devices particularly in handling hardware-software interactions and resource utilization.
• Sensor Fusion: Experience with camera localization and motion estimation using a combination of GPS IMU video and magnetometer.
• Core CV Skills: Experience with advanced computer vision techniques including camera localization motion estimation and 3D reconstruction.
• System-Level Software: Proficiency in low-level system software and hardware-software interactions on Qualcomm chipsets.

Additional Information :

Our Commitment

We are committed to building diverse teams and creating an inclusive workplace that enables all Grabbers to perform at their best regardless of nationality ethnicity religion age gender identity sexual orientation and other attributes that make each Grabber unique.

Benefits at Grab:

• Insurance: Comprehensive Term Life Insurance and Medical Insurance.
• Customized Benefits: GrabFlex offers a tailored benefits package.
• Parental Leave: Maternity and Paternity Leave for new parents.
• Support Programs: Confidential Grabber Assistance Programme for lifes challenges.
• Well-being Initiatives: Access to including health programs webinars and events.
• Work-Life Balance: FlexWork arrangements to support personal and professional life.

Remote Work :

No

Employment Type :

Full-time