Research Assistant - SADI for Continuous AI Processing

The postholder will support the development and validation of SADI (Sum-Difference Accelerated Asymmetric Inner Product), an energy-efficient FPGA architecture for onboard AI inference in small satellites. The role will focus on translating the proposed architecture into a parameterisable VHDL implementation and evaluating its performance against conventional FPGA-based neural-network accelerators.

Key activities will include:

• RTL/VHDL design
• FPGA synthesis and implementation
• Hardware-level performance evaluation
• Simulation of quantised neural-network workloads relevant to satellite Earth-observation applications

The researcher will contribute to benchmarking the architecture in terms of power consumption, logic utilisation, throughput and computational efficiency, with particular emphasis on SAR-based ship detection and, where appropriate, multispectral image segmentation. The role will also involve integration and testing on the University's existing FPGA hardware, analysis of results, and preparation of technical documentation, architecture specifications and research outputs.

The postholder will work closely with the project lead and relevant research and industry stakeholders to assess the potential for translation of the technology towards space-qualified FPGA platforms and future CubeSat applications. The position provides an opportunity to develop practical expertise at the intersection of AI hardware acceleration, FPGA design, embedded computing and space technology, while contributing directly to the project's intended knowledge-transfer and impact objectives. The researcher will also support dissemination of the project's outcomes through conference participation, publications and the release of an open-source reference implementation where appropriate.

Education, Qualifications and Experience

Essential Criteria

• BEng/MEng/PhD (or equivalent) in Computer Engineering, Electronic/Electrical Engineering, Computer Science, or a related discipline.
• Demonstrable experience in deploying neural networks/AI models on embedded, edge, FPGA, or other hardware-constrained platforms.
• Availability to commence by 1 October 2026.

Primary skills:

• VHDL/RTL design
• FPGA development
• Digital hardware design
• Neural-network acceleration
• Hardware simulation and performance analysis

Heriot-Watt University is committed to securing equality of opportunity in employment and to the creation of an environment in which individuals are selected, trained, promoted, appraised and otherwise treated on the sole basis of their relevant merits and abilities. Equality and diversity are all about maximising potential and creating a culture of inclusion for all. Heriot-Watt University values diversity across our University community and welcomes applications from all sectors of society, particularly from underrepresented groups. We welcome and will consider flexible working patterns e.g. part-time working and job share options.