Python Fabric VS DeepMark

DeepMark's answer:

DeepMark annotates the script rather than returning a score. Every mark is evidenced with a comment placed on the exact sentence that earned it, on handwritten work as well as typed, so a teacher can audit the marking instead of trusting it. Most AI marking tools output a number and a paragraph of generic feedback, which a teacher has no way to check and a student cannot act on.

The second difference is that marking is collaborative. DeepMark works like a shared document, so a department can mark, moderate and standardise together in real time. Standardisation is the part of assessment that costs schools the most time and is hardest to evidence, and it is the part almost no AI marking tool addresses.

DeepMark's answer:

• The mark scheme is yours. DeepMark marks against the scheme you already use, including levels of response, rather than fitting your subject to a generic rubric. That is the difference between output a head of department can defend and output they cannot.
• You can check the work. Inline annotations mean the marking is auditable in the two minutes a teacher actually has, rather than requiring blind trust in a model.
• Built for UK secondary specifically. GCSE and A-Level extended writing, exam board mark scheme structures, UK data protection expectations and the DfE product safety standards for generative AI in education. Most tools in this category are built for the US market or for higher education and adapted afterwards.
• Priced for a teacher, not a procurement cycle. An individual teacher can sign up and mark a set tonight without asking anyone's permission.

DeepMark's answer:

UK secondary school teachers who mark extended writing: English, History, Geography, Business, Economics, Sociology, Politics and Religious Studies. These are the subjects where a single class set is several hours of marking and where the mark scheme is a levels of response grid rather than a list of correct answers.

Within that, the core user is a classroom teacher or head of department at Key Stage 4 and 5 who is marking mocks and coursework alongside a full timetable. Private tutors and tutoring agencies working with GCSE and A-Level students are a second audience, where marking turnaround is the product being sold rather than an overhead.

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.

DeepMark's answer:

• Next.js and React
• SST for infrastructure as code
• AWS (UK region)
• PostgreSQL (serverless)
• Real time collaborative document editing
• Provider-independent LLM layer
• Model Context Protocol (hosted MCP server)