In the Department for Education’s Areas of Research Interest report, AI is presented not as a silver bullet but as a tool with powerful promise — one that must be deployed with care, equity, and rigorous evaluation. My project to build a local LLM at my SEMH provision is very much in step with that vision. 

At the SEMH unit I head, I’m developing an AI‑backed system that helps staff capture structured daily observations of pupils’ social, emotional, and mental health (SEMH). We use voice notes (transcribed into text) against a unified checklist derived from the 6 Core Strengths and Trauma Perceptive Practice (TPP) frameworks. Over time, these observations feed into the local LLM, which can detect patterns, generate weekly reports, support‑plan drafts, and 24‑week progress reviews. Importantly, everything is local — the LLM only works on data we intentionally load (dummy data to start with) — to maximise privacy control. 

Where Our Project Aligns with the DfE’s AI Agenda 

DfE Priority / Research Goal	How Our LLM Project Contributes	Notes / Caveats
Support child development and wellbeing	By modelling pupils’ SEMH trajectories, our system can surface trends that staff might otherwise miss.	AI is an assistant, not a substitute for professional judgement.
Improve teaching and reduce workload	Automating drafting of progress reviews and analysis saves teachers’ time.	Human oversight must remain central.
Inclusion & assistive technologies	Helps surface insights about pupils whose needs may be subtle or emerging.	Avoid reinforcing bias or overpathologising.
Productivity, data use & decision support	Proof of concept in better data capture, turning qualitative observations into insight.	Local and constrained, strengthening control.
Safety, ethics, equity, inequality	Designed with privacy, bias mitigation, and validation in mind.	Robust validation and stakeholder engagement essential.
Evidence & evaluation	Track AI outputs against independent human assessments.	Generates evidence for larger trials.

This project reflects the DfE’s invitation to test how AI can support education, provided it is done ethically, equitably, and rigorously.