PhD Studentship: Sovereign Agentic AI: Distilling Small Language Model Agents for Local Deployment

Overview

This doctoral studentship focuses on sovereign agentic artificial intelligence, investigating how to distil small language model agents for local deployment. The project addresses the reliance of UK businesses on overseas cloud providers for large model agents by exploring whether small language models running on local workstations or on-premises servers can serve as reliable agents for bounded business workflows such as procurement, invoice processing, and customer support. The work involves collaboration with an industry research lab and is supervised by Professor Shiqiang Wang and Dr. Tianjin Huang within the Department of Computer Science at the University of Exeter.

What you get

The studentship provides coverage of tuition fees and a stipend of £21,805 per annum for the 2026/7 academic year. The recipient will join regular meetings with supervisors and an industry collaborator, have opportunities for possible visits to the industry research lab, and aim to publish research outcomes at leading machine learning venues.

Requirements

The project investigates two complementary routes involving harness optimisation and the fine-tuning of small language models on traces of successful agent behaviour derived from larger teacher models and tool execution feedback. The student will develop automated harness optimisation methods under fixed local computation and latency budgets, study the transfer of agent behaviours, and measure reliability, cost, latency, and data locality using realistic task suites while producing open tooling suitable for adoption by small and medium-sized enterprises.

Key dates

The closing date for applications is 2026-12-08.