England, Wales and Northern Ireland · Extended Project Qualification · Computing and AI artefacts
EPQ computer science artefact mentoring
A growing number of sixth formers want their Extended Project to be something they built: an app, a website, a game, a machine learning model that classifies images or predicts something from data. Both AQA and Pearson allow it. An artefact is a recognised form of EPQ product, provided it comes with a written research report. But the specifications are just as clear about whose work it must be. At AQA the supervisor signs that the evidence is "the unaided work of the student", and the biggest assessment objective, worth 40 per cent, is the student's own development of the product. So a mentor's job is narrow and real: teach the programming and the method, on separate practice material, so the student can build and evaluate the artefact themselves. This page explains how computing artefacts fit the EPQ, where the marks sit, the authenticity rules, and how our mentoring works.
Live teaching since 2020 · 10,000+ students · Built from the AQA 7993 and Pearson EPQ specifications
In short
An EPQ computer science artefact is an Extended Project Qualification (AQA 7993 or the Pearson Edexcel Level 3 Extended Project) whose product is something built, such as an app, website or AI model, with a research-based written report of at least 1000 words at AQA. The qualification is 120 guided learning hours, and at AQA the Develop and Realise objective carries 40 per cent. Modern Age Coders offers EPQ computer science artefact mentoring live online, one to one or in batches of 15 to 20, through its AI and Machine Learning for Teens course, teaching method on practice material and never writing or editing the project. The first session is free; we are not linked with any exam board or school.
Courses for EPQ builders
Three courses behind a computing or AI artefact
AI and Machine Learning for Teens suits model-based artefacts. Python for Teens is the foundation for any built project, and Full Stack Web Development covers app and website artefacts.
Yes. AQA accepts an artefact as the project product alongside a research-based written report of at least 1000 words, and Pearson runs a separate Artefact unit whose examples include creating a website. A program or a trained model fits naturally within both.
Computing artefacts at the two boards we read (specifications as published)
Point
AQA EPQ 7993
Pearson Edexcel Level 3 Extended Project
Size
120 guided learning hours
120 guided learning hours
Taught element
About 30 guided learning hours
About 40 guided learning hours recommended
Artefact route
An artefact plus a research-based report of at least 1000 words
Unit 4: Artefact, one of four units; examples include creating a website
Written-only route
A report of approximately 5000 words
Unit 1: Dissertation
Other evidence
Production Log and a presentation to a non-specialist audience
Activity log and a presentation
Grades
A* to E
A* to E
The AQA wording is exact: "Where the chosen product is an artefact there must also be a research based written report of a minimum 1000 words." It adds that artefacts can be evidenced with "photographs of each stage of development as well as the final product". For a computing artefact, the equivalent is screenshots, commit history, test records and short screen recordings at each stage, which also make the development visible to the supervisor. Every project also needs its title and aims approved by the school before the main work starts: AQA requires students to "draft a title and aims of the project for formal approval by the centre".
Why computing artefacts suit the EPQ
They produce evidence naturally: versions, tests and measurable results.
They suit students heading for computer science, engineering, data science or medicine, where a well-evaluated model shows the habits a degree will want.
Where they go wrong
A project that becomes a coding tutorial, describing what was built rather than investigating a question.
An ambitious build that is still unfinished in the spring, leaving nothing to evaluate. Both are avoidable with a sharp title and a realistic plan.
Sources: AQA, Level 3 Extended Project Qualification 7993 specification, version 1.4; Pearson, Level 3 Extended Project specification, Issue 4, November 2023. Both read on 31 October 2026. OCR also offers an Extended Project; we could not read its specification for this build, so it is not described here.
Where marks are
How is an EPQ artefact marked?
Against four assessment objectives. At AQA, managing the project is 20 per cent, using resources 20 per cent, developing and realising the product 40 per cent, and reviewing it 20 per cent. The artefact itself matters, but well over half the marks reward planning, research and evaluation around it.
AQA EPQ 7993 assessment objectives and what they mean for a computing artefact (weightings from the specification; the right-hand column is ours)
Objective
Weighting
In a computing or AI artefact
AO1 Manage
20%
A clear question, aims and a plan, monitored and adjusted in the Production Log
AO2 Use resources
20%
Research into existing methods and tools, critically selected and referenced, including any use of AI tools
AO3 Develop and realise
40%
The student's own design, build, testing and problem solving, with decisions taken critically
AO4 Review
20%
Evaluation against the original aims, own learning, and a presentation to a non-specialist audience
AQA describes AO3 as the ability to "Select and use a range of skills, including, where appropriate, new technologies and problem-solving, to take decisions critically and achieve planned outcomes." That is the objective most at risk when someone else helps with the build, because it is precisely the student's own skills and decisions that it measures.
The supervisor's checklist for approving a proposal asks whether the title lets the student "plan, research, analyse, evaluate and explain, rather than simply describe and narrate". For computing projects, that one test decides a great deal. "Building a chatbot" describes; "How accurately can a small model trained on a few hundred examples answer questions about one topic, and where does it fail?" investigates.
Ask a question
Frame the artefact as a test of something: accuracy, usability, speed or fairness, measured against a stated aim.
Plan for evaluation
Decide before building how success will be measured, and keep test data aside from the start.
Keep the log honest
Record dead ends and changes of plan. AO1 and AO4 reward the reasoning, not a smooth story.
Own work
Can someone help with my EPQ artefact?
Your supervisor guides you, and you can learn skills from anyone. But AQA requires the supervisor to declare that the evidence is the unaided work of the student, treats submitting work that is not your own as malpractice, and Pearson requires a signed declaration on the Candidate Record Sheet. A mentor may teach; a mentor may not build, write or fix the project.
AQA's specification sets out the position. Supervisors sign "a declaration that the evidence submitted for assessment is the unaided work of the student". Where extra help was received and is acceptable, "the supervisor should award a mark which represents the student's unaided achievement", and "If the supervisor is unable to sign the authentication statement for a particular candidate, then the student's work cannot be accepted for assessment." Students must not "submit work which is not their own" or "submit work typed or word-processed by a third person without acknowledgement", and AQA states: "These actions constitute malpractice, for which a penalty (eg disqualification from the qualification) will be applied."
Pearson's specification asks each learner to sign a declaration on the Candidate Record Sheet, warns that without confirmation "the learner will be marked as absent for the project", and on artificial intelligence says that, "Following the 2023 publication by JCQ of two guidance documents relevant to this qualification, centres are advised to support candidates with referencing and recording of any use of AI in relation to the candidates assessed work". Any AI tool used in the project must be declared and referenced as your school instructs.
Our mentoring line, drawn from the AQA and Pearson rules
Inside our lessons
Outside our lessons, always
Teach Python, data handling, machine learning and web skills in lessons
Write, debug, refactor or review the code of the artefact
Teach evaluation methods on public practice datasets
Touch, clean or analyse the student's project data
Explain what makes a question analytical rather than descriptive, with our own examples
Suggest, draft or rewrite the student's title, aims or plan
Teach how to reference sources and record tool use
Write, edit or proofread the report, log or presentation
Encourage students to tell their supervisor about outside lessons
Work against a school's instructions on outside help
Tell your supervisor
If you take lessons alongside an EPQ, say so in a supervisor meeting and note it in your log. A sentence such as "I am taking general Python lessons outside school, on separate practice exercises" is honest and protects you.
Your supervisor can then judge, as AQA asks, what represents your unaided achievement.
Age-appropriate tools
Online AI services set their own age rules in their terms of use. We teach with open libraries such as scikit-learn on public datasets, which need no account.
If your project uses any external AI service, check its age terms and your school's policy first.
Sources: AQA EPQ 7993 specification, version 1.4, sections on supervision, authentication and malpractice; Pearson Edexcel Level 3 Extended Project specification, Issue 4, sections on authentication and use of AI. Read on 31 October 2026.
Method
How do you evaluate a machine learning EPQ artefact?
With data the model has never seen, a comparison against a simple baseline, and more than one measure. Accuracy alone can flatter a model. Here is the method as we teach it, on an example of our own that is not anyone's project.
Example (ours). Imagine a model that labels garden photographs as "robin" or "not robin". It is tested on 120 photos kept aside from training: 48 contain a robin and 72 do not. The model finds 40 of the 48 robins (true positives), misses 8 (false negatives), wrongly labels 10 other photos as robins (false positives), and correctly rejects 62 (true negatives).
Our worked evaluation for the robin example
Measure
Calculation
Result
What it tells you
Accuracy
(40 + 62) ÷ 120
85%
Share of all photos labelled correctly
Precision
40 ÷ (40 + 10)
80%
When it says robin, how often it is right
Recall
40 ÷ (40 + 8)
About 83%
Of the real robins, how many it found
Baseline
Always answer "not robin": 72 ÷ 120
60%
The score a model with no skill would get
The baseline is the line most EPQ evaluations miss. A model at 85 per cent sounds strong until you notice that guessing "not robin" every time scores 60. The honest claim is that the model adds 25 percentage points over doing nothing, finds about five robins in six, and is wrong one time in five when it says robin. That is an evaluation an examiner can follow, and it leads naturally into the review: which photos fooled it, and why?
Practice code (ours) on a public toy datasetPython
# Practice only (ours): evaluate a classifier on a public toy dataset,
# never on the student's own EPQ data or code.
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import confusion_matrix, accuracy_score
X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.3, random_state=1, stratify=y)
model = DecisionTreeClassifier(max_depth=3, random_state=1).fit(X_train, y_train)
pred = model.predict(X_test)
print(confusion_matrix(y_test, pred))
print("accuracy:", round(accuracy_score(y_test, pred), 3))
We teach evaluation on public datasets such as this one. Students then apply the method to their own project themselves.
Split first. Put test data aside before any training, and never tune the model on it.
Set a baseline. The simplest possible rule, measured on the same test data.
Report several measures. Accuracy, precision and recall, or the measures that fit the question.
Look at failures. Examine the cases the model got wrong; they are the heart of the review.
Tie back to the aims. Did the artefact achieve what the approved aims said, and what would you change?
Try this (ours, with answers). A model is tested on 200 emails, 30 of them spam. It flags 25 emails, of which 20 are spam. (1) What is its precision? (2) Its recall? (3) What accuracy would "never flag anything" achieve? Answers: (1) 20 ÷ 25 = 80 per cent; (2) 20 ÷ 30, about 67 per cent; (3) 170 ÷ 200 = 85 per cent, which is why accuracy alone would mislead here.
Timeline
When should I start learning the skills for a computing EPQ?
Before the build starts, ideally before the proposal is approved. Learning a language and the evaluation method while also running the project leaves little time for the planning and review that carry 40 per cent of the marks at AQA.
A typical skills timeline alongside a one-year EPQ (ours; your school sets the real deadlines)
Stage
Project work (student and supervisor)
Skills lessons (with us, separate practice)
Before proposal
Choosing an area, reading around it
Python foundations, data files, simple models or web basics
Proposal and approval
Title and aims approved by the school
Evaluation method and baselines on practice datasets
Research
Sources, existing tools and methods, referenced
Reading documentation and technical papers, referencing practice
Build
The student designs, builds and tests the artefact
Topic lessons only, on unrelated exercises
Review and presentation
Evaluation, report, presentation to a non-specialist audience
Explaining technical ideas plainly, practised on our own topics
Students also taking A level Computer Science will find that the same rules shape its coursework; the A level Computer Science project page covers that separate project, and A level Computer Science tuition the course. Under AQA's rules, no work assessed for the EPQ may also be submitted for another qualification, so the two projects must be genuinely different.
Our mentoring
How does your EPQ artefact mentoring work?
It is skills teaching, timed around the project and kept apart from it. Students learn on our exercises and public datasets, then do their own project work with their school supervisor.
A free first session. We talk about the kind of artefact you are considering and where your programming stands, and agree a skills plan. We do not look at project documents.
Foundations. Python, data handling and testing, or web development, depending on the artefact type.
Method. Train and test splits, baselines, measures, user testing for apps, and how to record decisions.
Communicating. Explaining technical work to a non-specialist audience, practised on our own topics.
Boundaries throughout. No project code, data, title, report or slides are ever shared with or produced by us.
Shared batch or private sessions?
A batch of 15 to 20 sixth formers works well for building programming and machine learning foundations steadily.
Private sessions fit a student who needs one specific skill quickly, such as web back ends or model evaluation, before their build begins.
What we will not claim
Any EPQ grade, any university offer, or any link with AQA, Pearson, OCR or a school.
That our teaching replaces the school supervisor. The supervisor guides the project; we teach general skills.
Progress
Four rungs before the build
Skills a student should hold securely before the project build begins.
Our progression for EPQ computing students
Rung
Target time
Skill held without help
1. Programming
Year 12 autumn
Write, run and test small Python programs with files and functions
2. Data or web skills
Year 12 winter
Load and clean a dataset, or build a simple web app with stored data
3. Evaluation method
Before approval
Split data, set a baseline, choose measures, or plan user testing
4. Explaining
Before presenting
Describe a technical result to someone with no computing background
Rung three protects the review
A project with no planned evaluation struggles in AO4, however good the build.
Weekdays or weekends, flexible: choose your batch, or for a very personalised plan choose 1-on-1 at your convenience, weekend mornings or weekdays. Sessions are led from India yet booked for UK sixth formers on the UK clock.
After school
Around sixth-form timetables.
Evening
For students with free periods used for study.
Weekend
Longer blocks for a full practice exercise.
Skills, not projects
Lessons teach methods on our material, never on the EPQ itself.
Public datasets
Practice evaluation on open data that needs no account.
Baselines first
Every model we train in class is compared with doing nothing.
Plain explanation
Practice in presenting technical results to non-specialists.
Supervisor-friendly
Students are encouraged to log outside lessons openly.
Two formats
A shared batch of 15 to 20, or private one-to-one time.
Student work
Class builds by sixth formers and younger teens
Class projects only, none submitted for any qualification. More are in student labs.
AI and ML
NutriLife
An AI nutrition coach that reads what you eat and works you toward a target.
A weather forecasting site with live conditions for any location.
by Krish
Fees
Fees for EPQ skills sessions
The first session is free. Then one monthly fee, cheaper in a batch than one to one, begins only once a course and a weekly time are fixed. You can stop at any time.
Google reviews from families, quoted without change.
★★★★★
"The one step solution for my son. Modern Age Coders make learning coding so simple that kids love it. The teachers explain complex concepts clearly with practical exercises and interactive content."
Ria Mukherjee
Parent
★★★★★
"Modern Age Coders has been a game-changer for me. I struggled to grasp IT concepts and coding before joining, but their classes transformed everything. I can now confidently write complex programs with ease."
Samriddha Mondal
Student
★★★★★
"One of the most wonderful education centres out there. Education is not limited to school syllabus but focuses on skill development."
Vansh Agarwal
Student
★★★★★
"My child Dhairya is really enjoying the Modern Age Coders classes. This is his first online class and he eagerly looks forward to it. I can already see his improvement, and the teachers are very cooperative."
Sonam Oswal
Parent of Dhairya
★★★★★
"Modern Age Coders have wonderful teachers who teach in a clear, easy and practical way. The teacher boosts students' confidence and inspires them to learn without hesitation."
Sonu Goyal
Parent
★★★★★
"I highly recommend this computer coding class! The teachers are incredibly knowledgeable and passionate about coding."
Ritu Kedia
Parent
EPQ questions
What students ask about computing and AI EPQs
Can my EPQ be a coding project?
Yes. AQA accepts an artefact as the product alongside a research-based written report of at least 1000 words, and Pearson has an Artefact unit whose examples include creating a website. An app, game or machine learning model fits either.
How long is the written report for an artefact EPQ?
At AQA, an artefact needs a research-based written report of a minimum of 1000 words. A project that is only a written report should be approximately 5000 words.
How is an EPQ marked?
At AQA against four objectives: Manage 20 per cent, Use Resources 20 per cent, Develop and Realise 40 per cent, and Review 20 per cent. Grades run from A* to E.
What are good EPQ ideas for computer science?
Strong ideas ask a question the artefact can test, such as how accurately a small model classifies something, or whether an app design improves a task for real users. Your supervisor must approve the title and aims, so we never suggest yours.
Can a tutor help with my EPQ?
A tutor can teach general skills on separate practice material. AQA requires the supervisor to declare the evidence is the unaided work of the student, and treats submitting work that is not your own as malpractice, so nobody else should write, build or edit the project.
Can I use AI tools in my EPQ?
Only as your school allows, and any use must be recorded and referenced. Pearson advises centres to support candidates with referencing and recording of any use of AI, following JCQ guidance published in 2023.
How many hours is an EPQ?
Both AQA and Pearson specify 120 guided learning hours, including a taught skills element: about 30 hours at AQA and about 40 recommended at Pearson.
Can my EPQ overlap with my A level Computer Science project?
No. AQA says no work assessed for the EPQ may be submitted for another accredited qualification, so the two projects must be different pieces of work.
When are sessions?
Weekdays or weekends, flexible: choose your batch, or for a very personalised plan choose 1-on-1 at your convenience, weekend mornings or weekdays. Every time is set in UK time.
Are you linked with AQA, Pearson or my school?
No. We are a separate online teaching business, unconnected to AQA, Pearson, OCR or any school or college.
Related pages
Other pages for sixth-form programmers
Coursework rules elsewhere, and where AI skills lead.
Share a phone number and pick a UK time for our callback. The session is a conversation about your artefact type and your programming, ending with a skills plan that stays separate from the project.
WhatsApp is free from a UK mobile and usually quickest. Our team answers from India, on a +91 number.
Want a class sooner? Book a priority demo
A priority demo is a full class of about 45 to 60 minutes. After the class the mentor sends a written skill report and a personal learning roadmap on WhatsApp. If you enrol, the demo fee is adjusted against your first month's fee. Share your demo Order ID with us on WhatsApp and we take it off.
