DeepMark's answer:
DeepMark was founded by a teacher and a software engineer who both grew up around education.
Geoff Waugh has spent over twelve years in education, first as a university lecturer and now as a GCSE Business Studies teacher. As a single father, he watched weekends and family time disappear into marking. He did not encounter teacher burnout as a policy abstraction. He lived it. Stuart Bourhill is a senior software engineer of more than a decade, from a family of teachers, who recognised that AI had advanced well beyond anything schools could realistically access.
The numbers behind the problem are stark. Over 114,000 teachers left the profession in three years. One in three of those still teaching are considering leaving within the next twelve months, and nine out of ten of them name workload as the reason. Not pay, not pupils, the weight of work that follows them home. Marking is the single most time-consuming non-teaching task in the profession and the one most resistant to the support schools can realistically offer. Teachers do not need more wellbeing initiatives, they need less work.
When the two of us reconnected, the conversation was straightforward: the technology finally existed to give teachers their time back. What was missing was a product built for the realities of a classroom. The tools available were inconsistent, unsafe for sensitive student data, and disconnected from how schools actually work. They were built by people who understood software rather than people who understood classrooms.
We also realised AI makes something possible that no teacher could ever do with a red pen. A view across an entire cohort. Where a class is consistently misunderstanding a concept, which question types produce the most errors, which feedback patterns correlate with better outcomes over time. Not just less work for teachers, but better data for schools.
Four things we will not move on. Professional judgement stays with teachers, every mark is provisional until a teacher confirms it. Safeguarding is non-negotiable, student data is never used to train external models. Full transparency on every mark, each one links to the exact passage that earned it. Built around real classroom workflows, not a generic AI tool adapted for schools after the fact.
Supabase
Android Studio
AppWrite
Back4App
Xcode
Socket.io
Vercel
excelas.ai
Edexia
iki.ai
Python Fabric
Heyday
recALL
