AI will either build human capability or quietly erode it. That isn't a prediction — it's a design decision, and almost nobody is making it on purpose.¹

Fifteen years engineering education that demonstrably works — curriculum, assessment, and the unglamorous systems underneath. I treat AI as the next great research science, and applied neuroscience as the discipline that decides what it does to us.

I build the thing, then I write down what happened — at The EdJournal. Currently finishing an MSc in Applied Neuroscience at King's College London.

the whole site is basically
this one sentence

§1 Selected work

Fig. 1 — Flagship · iterated across four model generations

AI by Design

A year-long body of work on AI in education: a framework, a position paper, a state-of-use review — culminating in a complete K–12 AI curriculum, designed end to end.

Fig. 2 — Research · adversarial, on purpose

Stress-testing AI on high-stakes assessment

Trained an AI marking system on real exam papers and anonymised student responses, then went looking for its biases. Finding out where it breaks is the finding.²

Fig. 3 — Shipped · in classrooms, not slide decks

AI teaching assistants

A chemistry assistant producing the full teaching content stack — lesson plans, differentiated tasks, guidance through complex practical work. Built, used, iterated.

the marking bias
turned out to be
the interesting part

§2 Observations

350+

practitioners trained across a multi-site trust

70%

of teacher planning time removed via AI workflows

4

continents reached by a venture built from nothing

80%

of a 250-student cohort met or beat targets, two years running

1

TEDx talk arguing grades don't matter. I stand by it.

every number here
has a story where
it nearly didn't work

§3 Where I've been

Founder & WriterThe EdJournal
2023 —
Head of Science & Curriculum LeadershipInternational School of Zurich North, Switzerland
2023–25
Cohort Lead & IB DP CAS CoordinatorInter-Community School Zurich
2020–23
Cohort Lead & Team ManagerSt. John's International School, Belgium
2018–20
Founder & DirectorThe GradeLab — scaled across four continents, entirely by word of mouth
2016–18
Director of Teaching & LearningSouthmoor Academy Trust, UK — 350+ practitioners
2013–15
Junior ChemistAkzoNobel — Interpon Powder Coatings³
2007–09

accidentally became
an instructional designer
at a paint company

§4 Grounding

MSc Applied NeuroscienceKing's College London
in prog.
MA International EducationUniversity of Bath
2023
Fellow, Chartered College of TeachingFCCT · and MRSC, Royal Society of Chemistry
—
TEDx SpeakerGrades Don't Matter; Wellbeing and Life Skills Do
2020

§5 Elsewhere

The EdJournalReimagining education one bold idea at a time.
publication
Vanguard Education GroupA house of education ventures.
ventures
LinkedInThe tidier version of all this.
profile

§6 Get in touch

Advisory, speaking, collaboration — or an argument about where AI and learning are actually going. I'd rather have the argument. shilpakapur@outlook.com

Open to the interesting

1.Deliberately, anyway. Plenty of people are making it by accident, at scale, right now.

2.The model was confidently wrong in a beautifully consistent pattern. That pattern is now a human-in-the-loop check.

3.Powder coatings. Genuinely fascinating. Nobody believes me.