Research Assistant or Associate in AI Model Optimisation for Edge Devices & NVIDIA Holoscan - Imperial College London

Job number

ENG03990

Faculties

Faculty of Engineering

Departments

Department of Electrical and Electronic Engineering

Salary or Salary range

£45,399- £59,484 per annum

Location/campus

South Kensington Campus - Hybrid

Contract type work pattern

Full time - Fixed term

Posting End Date

13 Aug 2026

About the role

We are seeking a Research Assistant or Research Associate to work at the intersection of AI model optimisation, GPU kernel development and FPGA-based hardware integration. The project targets the seamless integration of computer vision FPGA-based IPs with NVIDIA's Holoscan Sensor Bridge — a cutting-edge technology enabling low-latency, high-throughput streaming between sensors and edge GPU platforms.

The project involves Imperial College London and an industrial partner, Heronic Technologies (https://www.heronic.ai), aiming to revolutionise the “Sense-Decide” pipeline in edge automation.

You will contribute to building a system that tightly couples custom FPGA-based AI-ISP accelerators with NVIDIA's GPU-powered edge platforms, with a focus on minimising latency while maintaining high performance and scalability. A significant part of the work will involve AI model optimisation and the customisation of edge GPU kernels to push system performance to its limits.

This is a genuinely multidisciplinary challenge, spanning AI model design, low-level GPU kernel engineering, and hardware-software co-design — an opportunity to advance the state of the art in how AI signal processing systems are built and deployed.

What you would be doing

• Investigating and developing system architectures that demonstrate low-latency, easy integration of custom AI-ISP accelerators with GPU platforms via NVIDIA's Holoscan Sensor Bridge
• Developing and evaluating the full system under object detection applications, assessing performance across latency and detection accuracy metrics
• Implementing models in machine learning frameworks (e.g. PyTorch) and applying hardware-aware efficiency metrics to evaluate energy, memory, and latency trade-offs
• Contributing to research publications and presenting results at academic conferences
• Collaborating closely with Prof Christos Bouganis and the team at Heronic Technologies, who are developing the FPGA-based AI-ISP accelerator
• Helping to bridge the gap between academic research and industrial impact in energy-efficient AI

What we are looking for

• A strong background in GPU programming, machine learning, digital hardware design, computer engineering, applied mathematics, or a closely related field
• Experience with software engineering for scientific computing or machine learning (e.g. PyTorch), GPU programming and/or digital hardware design (e.g. Verilog)