Edge AI Research Scientist: Latency-Focused Inference

A*STAR RESEARCH ENTITIES; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

A*STAR RESEARCH ENTITIES is seeking a PhD-level researcher to lead optimization of AI systems for high-throughput sensing and edge deployment. You will drive research into QAT, pruning, distillation, and NAS to balance latency and accuracy across distributed cameras and sensors.

The role requires mastery of inference engines (TensorRT/ONNX/OpenVINO) and hands-on CUDA/OpenCL parallel computing, aiming for end-to-end AI pipelines from sensor input to decision output.

Qualifications

• Ph.D. in Computer Engineering, Computer Science, Electrical Engineering, or a related field with a focus on High-Performance AI.
• At least 3 years of experience in AI deployment and optimization.
• Deep understanding of AI Inference Engines (TensorRT, ONNX Runtime, OpenVINO).
• Mastery of Model Compression techniques (Pruning, Quantization, Distillation).
• Expertise in parallel computing with CUDA/OpenCL.

Responsibilities

• Lead research into optimization techniques to minimize latency (QAT, pruning, distillation).
• Design scalable AI deployment architectures for high-throughput data from multiple cameras and sensors.
• Conduct hardware-software co-design for deployment targets (Jetson, TensorRT, FPGA).
• Develop asynchronous data pipelines from image acquisition to final decisions.
• Establish rigorous performance profiling benchmarks for latency and memory footprint.
• Collaborate with System Integrator to integrate optimized models into factory software.

Skills

C++

Python

High-Performance AI

Model Compression

FP16/INT8 deployment

Education

Ph.D. in Computer Engineering / Computer Science / Electrical Engineering

Tools

TensorRT

ONNX Runtime

OpenVINO

CUDA

OpenCL

Job description

A*STAR RESEARCH ENTITIES is seeking a PhD-level researcher to lead optimization of AI systems for high-throughput sensing and edge deployment. You will drive research into QAT, pruning, distillation, and NAS to balance latency and accuracy across distributed cameras and sensors.

The role requires mastery of inference engines (TensorRT/ONNX/OpenVINO) and hands-on CUDA/OpenCL parallel computing, aiming for end-to-end AI pipelines from sensor input to decision output.