🎓 Capstone 🔬 Final Thesis Project ⏱ 375 hours · 12 ECTS

Capstone Project — course structure

A complete, syllabus-driven guide to the BCSAI Capstone / Final Thesis Project (PC-CSAI.4.M.A): its aims, learning objectives, teaching methodology, the full year-long timeline of phases, milestones and deliverables, the assessment rubric, and the rules that govern the written report and oral defense. Throughout, the APEX concept site is used as the worked example of a prototype-type capstone.

Bachelor in Computer Science & Artificial Intelligence · IE University · Academic year 2025–26

Jump to project timeline ↓ Open the syllabus PDF

01 Overview 02 Objectives 03 Methodology 04 Assessment 05 Project structure 06 Report & defense 07 Policies 08 Glossary 09 References

01 · Overview

What this course is

The Capstone Project is the culminating course of the degree. It is not a lecture course — it is a year-long, supervised research-and-build project that asks each student to integrate everything learned across the programme into one substantial, original piece of work. The purpose is to bridge undergraduate study and professional practice, contributing something meaningful to a field of computer science, data science or AI while producing a portfolio-grade artefact.

Course code

PC-CSAI.4.M.A

Credits

12.0 ECTS

Workload

375 hours

Sessions

60 (live online)

Year · Semester

4th · Sem 2

👤 Professor & coordination

Prof. Alexandre Anahory de Sena Antunes Simões — aanahory@faculty.ie.edu. Ph.D. in Mathematics (Universidad Autónoma de Madrid), B.Sc. in Physics (University of Lisbon); research at the Instituto de Ciencias Matemáticas and the Center for Automation and Robotics, Madrid. Office hours on request by email.

Area: Functional Group — Program Direction. A Capstone Evaluation Committee, headed by the capstone project coordinator, approves topics, project types and supervisor choices.

🧭 Your supervisor

Each student finds and is approved for an academic supervisor who monitors progress, ensures the work meets the criteria, and signs off the deliverables. Outside professionals are allowed with committee approval.

At least 4 programmed meetings over the project (in person or virtual).
A work plan — frequency and timing of check-ins — is agreed in the proposal.
Other faculty can answer ad-hoc questions but are not expected to read drafts.

How to read this page. Because a capstone has no fixed weekly lecture topics (the 60 live-online sessions are supervisor meetings and practice time), this guide is organised around what the syllabus actually defines: aims → objectives → methodology → assessment → the project timeline of phases, milestones and deliverables → the report & defense rules. Use the timeline in section 05 as your operational checklist.

🧱 Prerequisites & assumed background

There is no formal pre-requisite course, but the capstone assumes the full BCSAI toolkit: fluency in Python and version control, a working command of data wrangling, statistics and machine-learning fundamentals, and the academic-writing and literature-search skills built across earlier years. You should be comfortable reading peer-reviewed papers, citing in APA, and standing up a reproducible code project. If any of these feel shaky, close the gap in the fall — before the build phase compresses your time.

🗓 How to plan your 375 hours

Treat the workload as a budget, not a deadline scramble. A workable split is roughly fall ≈ 60 h (topic discovery, supervisor outreach, proposal, early reading), then spring ≈ 300 h across the three deliverables — say ~90 h literature review, ~110 h methodology & modeling, ~100 h drafting, results and revision — leaving ~15 h for the defense rehearsal.

Reserve ~6–8 h/week in spring and protect it like a class.
Back-plan from each deliverable date, not from April 30.
Log AI use and sources as you go — reconstructing them later is the classic time-sink.

01b · Choosing a path

Three project types

Early on you decide on a topic (from the Catalyst video series, the supervisor list, or your own idea) and one of the project types the syllabus allows. Whichever you pick, the work must still address a real data-analytics, computer-science or AI problem; make theoretical, methodological and empirical contributions; contribute something to the body of knowledge in a field; and end in a written report and oral defense. The type does not change whether you write a report — only what its main body emphasises and what counts as the central artefact.

Prototype

⚙️ Build & benchmark

Build a data model or software-application proof-of-concept, study its performance empirically, and compare it head-to-head with existing solutions in the market. The centre of gravity is the working artefact and its measured behaviour, but the syllabus is explicit that a prototype must still carry theoretical and methodological weight — you are not excused from a literature review or a defensible method just because you shipped code. The report documents the problem, the architecture and design decisions, the evaluation methodology, the results, and an honest benchmark discussion. The deliverable is roughly a 25–50 page report plus the software (in the appendix/repo, outside the page count).

Central artefact: a runnable PoC with a reproducible performance study
Main-body emphasis: system design, experimental setup, benchmarks vs. state of the art
Common mistake: a slick demo with no baseline comparison or ablation — "it works" is not evidence

Venture

🚀 Launch a venture

Launch your own venture project and argue, with evidence, that it can create, deliver and capture value. The two pillars are a detailed business plan (problem and market, value proposition, go-to-market, unit economics, competition, risks) and a proof-of-concept that shows the underlying technology actually works on a representative task. The report ties strategy to evidence: it must read as more than a pitch deck in prose, grounding claims in market data and the PoC's results rather than optimism. The strongest ventures quantify the problem and tie every strategic claim back to something demonstrated.

Central artefact: a business plan + a working technology PoC
Main-body emphasis: value creation/delivery/capture, market & competitor analysis, feasibility
Common mistake: all strategy and no technical proof — or a PoC with no credible path to value

Research

🔬 Investigate a question

A quantitative or qualitative investigation of a problem (social, business, analytical, etc.). You propose or develop an appropriate methodology to answer a sharp research question, identify or build a model or algorithm, run it on collected data where applicable, and write up in a fully APA-formatted research report following the classic literature review → methods → results → discussion/conclusion arc. The artefact is the argument: rigour of design, validity of analysis, and the strength of the contribution to knowledge are what matter most. This is the most academically demanding path — and the most natural fit for students aiming at graduate study.

Central artefact: an empirical study answering a clearly stated question
Main-body emphasis: hypotheses, study design, statistics/analysis, threats to validity
Common mistake: a vague question and post-hoc storytelling around whatever the data happened to show

Policy brief

📜 Inform a decision

The syllabus also lists a policy brief as an allowed form: a concise, evidence-led document that translates a technical or data-driven analysis into actionable recommendations for a decision-maker (a regulator, a company, a public body). It still rests on real analysis — you investigate an issue, marshal evidence and existing work, and argue for a position — but it is written for impact and clarity rather than for a specialist research audience. Confirm scope and expectations with your supervisor early, since this path is less commonly chosen than the three above.

Central artefact: an evidence-based brief with clear recommendations
Main-body emphasis: issue framing, options analysis, trade-offs, recommendation
Common mistake: opinion unsupported by data, or burying the recommendation

🧭 How to choose — and what stays the same

Pick the type that matches your strongest evidence: if you can build and measure, lean prototype; if you can frame and analyse, lean research; if you have a commercial thesis plus a demo, lean venture; if your contribution is a recommendation for a decision-maker, consider a policy brief.

Every type needs theoretical + methodological + empirical contributions.
Every type ends in a 25–50 page APA report and a ≤15-minute defense.
The type is locked in the short proposal — changing it later needs committee sign-off, so choose deliberately.

🏔 APEX as the worked example — a Prototype capstone

The APEX concept site models exactly what a prototype-type capstone looks like in practice: it frames a real problem (the fragmented multisport athlete), proposes an architecture (a five-layer orchestration stack), implements interactive proofs-of-concept (the zone calculator, workout simulator, environmental-normalization and race-forecaster models), and benchmarks the idea against existing platforms (TriDot, Zwift, Strava, TrainingPeaks). Map it to the capstone deliverables:

Research question / motivation "Can one orchestration layer unify a triathlete's fragmented device & data ecosystem into a single adaptive coaching loop?"

Methodology & model Physiological zone models, TSS/CTL/ATL/TSB load math, a heuristic EnviroNormalization model, race-split projection.

Prototype / proof-of-concept The interactive simulators and the live workout chart on the concept site.

Comparison with the market The platform-capability matrix (logging / planning / adaptive AI / virtual / smart gym / social).

02 · Learning objectives

What you should be able to do

The capstone is assessed against four families of competence. Together they describe a student who can frame a question, ground it in theory, execute it with data and models, and deliver it on time, in writing and out loud.

🔎 Research abilities

Develop a research question and carry out a qualitative or quantitative project end-to-end: identifying appropriate sources, navigating search engines and databases, producing useful computer- or data-science models, and communicating results and conclusions clearly and persuasively.

Shown by: a question that is specific and answerable; a search strategy that finds the right literature and data; models that genuinely address the question; and a write-up a non-specialist on the panel can follow.

📚 Theoretical proficiency

Identify and critically engage the core theoretical approaches relevant to the project's question(s) — situating the work within the existing body of knowledge rather than treating it in isolation.

Shown by: a literature review that compares and critiques sources rather than listing them, names the gap your work fills, and frames your method as a deliberate choice among alternatives.

📊 Analytical skills

Use large amounts of data in data-analytics or AI models; analyse and properly cite academic texts to draw well-supported conclusions about a specific topic.

Shown by: appropriate methods for the data, honest treatment of limitations and uncertainty, reproducible analysis, and conclusions that follow from the evidence — not from what you hoped to find.

🧩 Transversal skills

Time-management and organizational skills to run a large-scale project to a timeline; communicate ideas in writing and orally; build a collaborative relationship with the supervisor — respecting internal deadlines and meetings and incorporating feedback.

Shown by: deliverables hit on time, feedback visibly incorporated between drafts, meetings come prepared, and a confident, well-rehearsed defense. This is the competence the supervisor's process grade most directly measures.

03 · Teaching methodology

How the course runs

IE University's teaching method is collaborative, active and applied — students build their own knowledge while the professor and supervisor guide. For a capstone, the overwhelming majority of the workload is independent project work, scaffolded by supervisor meetings and a series of process workshops.

Individual studying / project work95%

≈ 285 hours — the core build/research effort, done independently with supervisor guidance.

Lectures / live sessions5%

≈ 15 hours — live-online supervisor meetings and process workshops. (Total contact-time accounting ≈ 300 h; full project effort is ≈ 375 h.)

🗣 Supervisor meetings & sessions

The 60 live-online "sessions" are videoconferences whose content each supervisor specifies in their own plan — typically technical: programming, optimization algorithms, data cleaning/acquisition, modeling. The first four are formal programmed meetings; the rest are practice/working time.

🛠 Process workshops

Six face-to-face workshops support the thesis process — separate from the technical supervisor meetings — covering how to structure a thesis, conduct a literature review, develop and assess models, visualise data and defend the work.

1 · Demystifying the thesis process
2 · Unpacking the literature review
3 · The pieces of the puzzle
4 · Model development & assessment
5 · Data visualization, statistics & deployment
6 · Defending the thesis

Mandatory workshop attendance. You must attend at least half of the workshops. Failing to do so results in a failing grade on both calls for the academic year.

AI policy. Generative-AI tools may be used for specific tasks with acknowledgment, but the final written thesis must be original and may not be AI-generated. Inappropriate use is academic misconduct and can fail the assignment or the course. All submissions are checked for plagiarism (Turnitin, GPTZero, etc.).

04 · Assessment

How you are graded

A panel of four judges — your supervisor, an external second reader, and two outside panelists — gives the final grade. The headline split is 60% written report / 40% oral presentation, with the supervisor's continuous-evaluation share folded into the report side.

Written report (supervisor + external second reader)60%

40% from an external second reader (quality of the written thesis) + 20% from the supervisor (final product & continuous evaluation of the collaboration).

Oral presentation / defense (two panelists)40%

Quality of the oral defense, split equally between the two panelists.

Written report — external reader

40%

What's assessed Quality of the written thesis: clarity of the research question, rigour of methods, soundness of results, and the strength of the discussion against the existing market/literature.

High mark looks like A sharp question; a critical, well-cited literature review; a method that clearly fits the question; results presented honestly with limitations; and a discussion that benchmarks against real alternatives. APA is clean throughout.

Low mark looks like A fuzzy question, a list-not-critique literature review, unjustified method choices, results without baselines or error bars, and a conclusion that overclaims. Sloppy citations read as carelessness.

How / by whom An external second reader, independent of the supervision.

Written report — supervisor

20%

What's assessed Split equally between the quality of the final product and the process — the collaboration, meeting attendance, and timeliness/quality of the three spring deliverables.

High mark looks like Deliverables on time and improving draft-over-draft, feedback visibly acted on, prepared meetings, and a finished thesis the supervisor is confident putting their name behind.

Low mark looks like Late or thin deliverables, ignored feedback, silence between meetings, and a last-minute scramble. This is the most controllable 20% in the whole grade — losing it is an own goal.

How / by whom The academic supervisor (continuous evaluation).

Oral presentation — panelist A

20%

What's assessed An up-to-15-minute presentation of research question(s), methodology, key findings, conclusions and implications, then ≤20 minutes of panel questions.

High mark looks like A clear narrative arc, on-time delivery, well-designed slides, and answers that show you understand why you did what you did — including graceful handling of "what would you do differently?"

Low mark looks like Reading slides verbatim, overrunning, and defensive or hand-wavy answers under questioning.

How / by whom First outside panelist on the defense panel.

Oral presentation — panelist B

20%

What's assessed Same criteria as panelist A — persuasiveness, command of the work, and quality of responses under questioning.

High mark looks like Consistent command of the work from a second independent viewpoint; the two panelist scores converge when the defense is genuinely strong.

Low mark looks like Answers that contradict the report, or evident gaps between what was written and what is understood.

How / by whom Second outside panelist on the defense panel.

How the weights combine. The final mark out of 10 is a weighted sum: report 60% (external reader 40% + supervisor 20%, the supervisor's share itself half product / half process) plus defense 40% (two panelists, 20% each). So four people score you across two artefacts, and no single judge can carry or sink the grade alone. Practically: the report is where most points live, the supervisor's 20% is the part most under your day-to-day control, and the defense is where a strong report is confirmed or undermined live. A great thesis with a weak defense — or vice versa — still leaves real marks on the table.

Pass & re-sit rules. If the written report is graded below 5, you must make the required changes to pass it. In the second call the maximum attainable grade is 8/10, and you must repeat the defense. Missing the proposal deadline (or failing resubmission) automatically fails the first call and forces a June second-call defense.

05 · Project structure

The project, phase by phase

The capstone runs across the whole academic year as a sequence of gated phases. The fall semester is about choosing and proposing; the spring semester is about building and writing, scaffolded by three supervisor deliverables; the year closes with the written report and oral defense. Each phase below lists its objective, what you must hand in, and how it is evaluated.

PHASE 0 · FALL

Orientation & topic discovery

Sep 22–23

Understand the capstone process and converge on a viable topic and project type. Watch the Catalyst video case series and complete the reflection form, which seeds emerging-technology questions you might pursue. In parallel, start sounding out supervisors early — the best-fit advisors fill their advisee slots fast. The purpose of this phase is breadth before commitment: it is cheap to explore many directions now and expensive to pivot once the proposal is approved. The common mistake is treating orientation as a formality and arriving at the short-proposal deadline with no supervisor lined up.

📦 Deliverable / milestone

Watch the Catalyst video case series
Complete the case-series questionnaire form (Sep 23)
Begin sounding out potential supervisors

🎯 Evaluation

Completion gate — not directly graded
Feeds the "process" component via supervisor engagement

PHASE 1 · FALL

Short proposal & supervisor

by Oct 13–14

Commit to a direction: secure an approved supervisor and submit the short proposal (via the dedicated Blackboard assignment) that lets the Evaluation Committee assess feasibility. The format is deliberately light — a topic with a brief description and tentative title, the project type, and the supervisor's name — but the choices are consequential: the type you name here is the one you are expected to carry, and the supervisor must be committee-approved (outside professionals are allowed but need explicit sign-off). Aim for a topic that is narrow enough to finish in 375 hours yet substantive enough to make a real contribution. The common mistake is an over-broad topic ("AI in healthcare") that cannot be scoped into a single defensible study.

📦 Short proposal (Blackboard)

Topic — brief description + tentative title
Project type (prototype / venture / research)
Name of supervisor (committee-approved)

🎯 Evaluation

Reviewed by the Capstone Evaluation Committee
Supervisor choice must be approved

PHASE 2 · FALL

Full proposal (gate)

Nov 14 · review Nov 15–24

Working in collaboration with your supervisor, refine the project and submit a formal ~500-word proposal via Blackboard. It must contain three things: a detailed description of the subject and project type, a preliminary overview of the project timeline, and a summary of an initial literature review citing at least five academic sources. This is the first hard gate — the committee reviews every submission, and approval here is what unlocks the spring build phase. If a proposal is judged insufficient you get exactly one resubmission, due Dec 11; miss the deadline or fail the resubmission and you automatically fail the first call and must defend in the June second call (capped at 8/10). Treat the five sources as a real mini-review, not placeholder citations — they become the spine of Deliverable 1.

📦 Full proposal (~500 words)

Detailed description of subject & project type
Preliminary project-timeline overview
Initial literature review — ≥ 5 academic sources

🎯 Evaluation & gate

Committee reviews all submissions
If insufficient: one resubmission by Dec 11
No approval ⇒ fail 1st call, defend in June 2nd call

PHASE 3 · SPRING

Literature review

Deliverable 1 · Jan 31

Ground the project in theory by submitting a literature-review draft to your supervisor (due Jan 31). The expected form is a critical synthesis — typically several thousand words — that compares and contrasts prior work, organises it by theme or approach rather than chronologically, and ends by naming the gap your project fills. Workshop 1 (Demystifying the thesis process) and Workshop 2 (Unpacking the literature review) are timed to support exactly this. Its timeliness and quality feed the supervisor's process grade, and it is the backbone of the report's literature chapter. The common mistake is an annotated bibliography in disguise: summarising each paper in turn without ever putting them in conversation or staking out where you sit.

📦 Deliverable 1

Literature-review draft submitted to supervisor
Establishes theoretical framing & sources

🎯 Evaluation

Timeliness + quality feed the supervisor's "process" share
Reviewed in supervisor meetings

PHASE 4 · SPRING

Methodology, algorithms & preliminary analysis

Deliverable 2 · Feb 27

Design the engine of the project and submit Deliverable 2 (due Feb 27): the methodology written up, a schematization of the algorithm(s), and a preliminary analysis of the data. This is where vague intentions become a concrete, defensible plan — what data, collected or sourced how, run through which model or method, evaluated against what baseline and metric. Workshop 3 (The pieces of the puzzle) and Workshop 4 (Model development & assessment) support it directly. The supervisor assesses rigour and feasibility here precisely because it is still cheap to course-correct. The common mistake is jumping to a complex model before the data is understood — the preliminary analysis exists to catch leakage, class imbalance, missingness and scope creep before they sink the final results.

📦 Deliverable 2

Methodology written up
Schematization of the algorithm(s)
Preliminary analysis of the data

🎯 Evaluation

Supervisor assesses rigour & feasibility
Contributes to continuous "process" grade

PHASE 5 · SPRING

Full draft for revision

Deliverable 3 · Mar 30

Produce a complete full draft (due Mar 30) — every chapter present, results and discussion in near-final form — so the supervisor can guide final revisions before submission. "Full" is the operative word: a draft missing its discussion or conclusion cannot get the feedback that matters most, and this is the last process checkpoint before grading begins. Workshop 5 (Data visualization, statistics & model deployment) supports turning raw results into clear figures and a defensible deployment story. Budget the month between this draft and the Apr 30 report deadline for revision, not for writing the parts you skipped. The common mistake is submitting a 60%-complete draft and burning the supervisor's most valuable feedback window on gaps you already knew about.

📦 Deliverable 3

Full draft of the thesis for supervisor revision
Results & discussion in near-final form

🎯 Evaluation

Last process checkpoint before grading
Supervisor feedback to be incorporated

PHASE 6 · FINAL

Written report submission & approval

1st call: Apr 30 → May 10

Submit the finished thesis for grading: 25–50 pages (excluding software, appendices and bibliography), 12 pt, double-spaced, Arial or Times New Roman, APA throughout. The report carries 60% of the final grade — external reader 40% + supervisor 20% — and all submissions are screened for originality with Turnitin, GPTZero and similar tools. First-call deadline is Apr 30 with approval by May 10; the second call runs May 30 → Jun 10. If the report grades below 5 you must make the required corrections to pass it. The common mistake is leaving APA formatting, the bibliography and the abstract to the final night — readers notice, and a careless reference list quietly costs credibility on an otherwise solid thesis.

📦 Written report

1st call deadline Apr 30 · approval by May 10
2nd call deadline May 30 · approval by Jun 10
25–50 pages, 12 pt, double-spaced, APA

🎯 Evaluation

External second reader (40%)
Supervisor — product + process (20%)
Below 5 ⇒ mandatory corrections to pass

PHASE 7 · FINAL

Oral defense

1st call: May 20–31 · 2nd: Jun 20–30

Defend the work before the panel in a ≤15-minute presentation followed by up to 20 minutes of questions; the two outside panelists score it at 20% each, for 40% of the final grade. Cover the research question(s), methodology, key findings, conclusions and implications — and rehearse to time, because overrunning eats into your own Q&A and signals poor preparation. Workshop 6 (Defending the thesis, early May) is built for exactly this. Anticipate the obvious questions ("why this method?", "what are the limitations?", "what would you do with more time?") and prepare honest answers; the panel is testing command of the work, not trying to trap you. The common mistake is reading dense slides aloud instead of telling the story of the project. In a second call the defense is repeated and the maximum grade is 8/10.

📦 Presentation

≤ 15 min presentation; ≤ 20 min panel Q&A
Cover question(s), methods, findings, conclusions, implications

🎯 Evaluation

Two outside panelists (20% each = 40%)
2nd call: defense repeated, max grade 8/10

05b · Calendar

Full project calendar

Every dated milestone, workshop and deliverable in one table. Green rows are graded/process deliverables; red rows are hard gates; workshops are tagged WS.

Milestone	Type	Date
Catalyst video case-series questionnaire	Orientation	Sep 22–23
Short proposal deadline DL	Proposal	Oct 13–14
Full proposal deadline GATE	Proposal	Nov 14
Committee review of proposals	Review	Nov 15–24
Proposal resubmission (if not approved) GATE	Resubmission	Dec 11
Workshop 1 · Demystifying the thesis process WS	Workshop	January
Workshop 2 · Unpacking the literature review WS	Workshop	January
Deliverable 1 · Literature-review draft DL	Deliverable	Jan 31
Workshop 3 · The pieces of the puzzle WS	Workshop	February
Workshop 4 · Model development & assessment WS	Workshop	February
Deliverable 2 · Methodology, algorithms & prelim. analysis DL	Deliverable	Feb 27
Workshop 5 · Data viz, statistics & deployment WS	Workshop	March
Deliverable 3 · Full draft for supervisor revision DL	Deliverable	Mar 30
Written report deadline — 1st call GATE	Report	Apr 30
Approval of written report — 1st call	Approval	May 10
Workshop 6 · Defending the thesis WS	Workshop	Early May
Oral defense — 1st call DEFENSE	Defense	May 20–31
Written report deadline — 2nd call	Report	May 30
Approval of written report — 2nd call	Approval	Jun 10
Oral defense — 2nd call DEFENSE	Defense	Jun 20–30

06 · Report & defense

The written report and oral defense

The two graded artefacts have concrete format requirements. The report structure below is the default unless your supervisor approves an alternative; the main body's sections vary with project type.

📄 Written-report structure — and what each part holds

Title page — title, your name, supervisor, degree, institution and date.
Acknowledgements — brief thanks; also the natural home for the AI-use acknowledgment statement.
Table of contents — auto-generated, with figures/tables listed.
Abstract — one standalone paragraph: problem, method, key result, conclusion. Written last, read first.
Introduction — the problem, why it matters, your research question(s), and a roadmap of the report.
Main body — sections by project type (orientative): literature review, methodology, data collection, experimental design, business plan, results. This is where the type shows — a prototype foregrounds design & benchmarks, research foregrounds method & analysis, venture foregrounds the business plan.
Discussion & conclusion — a synthesis showing comprehensive understanding, interpreting results, stating limitations, and explicitly comparing your technology/results with existing market solutions.
Bibliography — APA, complete and consistent; excluded from the page count.
Appendix (if relevant) — extra tables, derivations, prompts, extended figures.
Software (if applicable) — code/repository; excluded from the page count.

The structure is the default unless your supervisor approves an alternative. Only the main body and the discussion count toward the 25–50 page limit.

🎙 Format & defense rules

Length — 25 (min) to 50 (max) pages, excluding software, appendices and bibliography.
Typography — 12 pt, double-spaced, standard margins; Arial or Times New Roman.
Citations — APA formatting required.
Integrity — checked with Turnitin, GPTZero and similar tools.
Presentation — up to 15 minutes by the student.
Questions — up to 20 minutes from the panel.
Cover — research question(s), methodology, key findings, conclusions and implications.

🎯 What the panel looks for

The defense is not a re-reading of the thesis; it tests whether you own the work. Panels reward a clear narrative, evidence of independent judgement, and honest engagement with limitations.

A story, not a summary — problem → question → method → findings → so-what, in 15 minutes.
Command under questioning — why this method, why this data, what you'd change.
Honesty about limits — naming a weakness before they do builds credibility.
Time discipline — finishing on time leaves room for the Q&A you're scored on.

🛠 A defense-prep checklist

Rehearse the full talk to time, out loud, at least twice.
Prepare 3–5 backup slides for likely questions (limitations, alternative methods, future work).
Know your numbers cold — every figure on a slide is a question waiting to be asked.
Have a one-sentence answer to "what is your contribution?"
Practise saying "I don't know, but here's how I'd find out" gracefully.

The defense panel. Four judges: your supervisor, an external second reader, and two outside panelists. The written report (60%) is judged by the supervisor and second reader; the oral defense (40%) by the two panelists. The two panelists may not have lived with your project the way your supervisor has, so the presentation must stand on its own — assume an informed but non-specialist audience.

07 · Policies

Rules & policies

The university-wide policies apply, and the Program Director may add indications. The capstone-specific re-sit rules and the AI policy are the ones most likely to bite — read them carefully.

🔁 Re-sit / re-take

Report below 5 ⇒ make required changes to pass it.
Second-call maximum grade is 8/10.
Second call requires repeating the defense.

🤖 AI policy

GenAI allowed for specific tasks with acknowledgment; the final thesis must be original and not AI-generated. Misuse is misconduct. A suggested AI-acknowledgment statement is provided in the syllabus.

📋 Conduct · attendance · ethics

Follow the University's Code of Conduct, Attendance Policy and Ethics Code; ≥ half of the workshops are mandatory. The Program Director may provide further indications.

08 · Key concepts

Glossary

Capstone, research-methodology and project-management terms you will meet throughout the year, plus the domain terms from the APEX worked example.

Capstone project: The culminating final-thesis project that integrates and synthesizes skills from across the degree into one substantial, original work.
ECTS credit: European Credit Transfer System unit of study workload; this course is 12 ECTS, ≈ 375 hours of student effort.
Supervisor: The approved academic mentor who guides the project, ensures it meets criteria, and grades 20% (product + process).
Evaluation Committee: Body headed by the capstone coordinator that approves topics, project types and supervisor choices, and reviews proposals.
Second reader: External examiner, independent of supervision, who grades the written report (40%).
Defense panel: Four judges (supervisor, second reader, two outside panelists) who decide the final grade.
Short proposal: Early submission stating topic, project type and supervisor for committee feasibility review.
Full proposal: ~500-word formal proposal: detailed subject description, preliminary timeline, and an initial literature review of ≥ 5 sources.
Deliverable: One of three spring submissions to the supervisor (lit-review draft, methodology, full draft) feeding the process grade.
Milestone: A dated, checkable point in the project calendar — proposals, deliverables, deadlines, defenses.
Gate: A milestone whose approval is required to proceed; failing it can force a second-call defense.
Continuous evaluation: Ongoing assessment of the collaboration, timeliness and quality of work — the supervisor's "process" share.
Prototype project: Project type building a model/software proof-of-concept, studying its performance and benchmarking it against the market.
Venture project: Project type launching a venture with a business plan (value creation/delivery/capture) and a proof-of-concept.
Research project: Project type investigating a question with a methodology, model/algorithm and data, written up APA-style.
Literature review: Critical survey of prior work that situates the project within the existing body of knowledge.
Methodology: The systematic approach — methods, data, algorithms — used to address the research question.
Proof-of-concept (PoC): A minimal implementation demonstrating that an idea or technology works in practice.
APA formatting: American Psychological Association citation and document style required for the thesis.
Abstract: A concise standalone summary of the question, methods, results and conclusions.
Plagiarism check: Automated originality screening (Turnitin, GPTZero) applied to all submissions.
First / second call: The two examination windows; the second call caps the grade at 8/10 and repeats the defense.
Catalyst video series: IE's case-series videos used during orientation to seed topic ideas and emerging-tech reflection.
Impact Xcelerator: IE research labs students are encouraged to collaborate with when choosing a topic.
Orchestration platform: (APEX) A layer that unifies many devices, protocols and services into one coherent system — the prototype's core idea.
TSS · CTL · ATL · TSB: (APEX) Training-load metrics — Training Stress Score and Chronic/Acute load and Stress Balance — modeling fitness, fatigue and form.
Periodization: (APEX) Structuring training across nested timescales (macro/meso/microcycle) — analogous to phasing a project plan.
Adaptive coaching loop: (APEX) A closed loop where each completed workout re-plans the next — the prototype's differentiating mechanism.
Benchmarking: Comparing your artefact's results against existing market/state-of-the-art solutions — required in the discussion.
Policy brief: An allowed project form: a concise, evidence-led document translating analysis into actionable recommendations for a decision-maker.
Research question: The single, specific, answerable question the whole project exists to address; everything else hangs off it.
Hypothesis: A testable statement predicting a relationship or outcome, evaluated against evidence in a research-type project.
Scope: The deliberate boundary of what the project will and won't cover — set early to keep 375 hours achievable.
Scope creep: The gradual expansion of a project beyond its agreed scope; a leading cause of missed deliverables.
Baseline: A reference method or result your work is measured against; "better than nothing" is not a baseline.
Ablation study: Systematically removing components of a model to show which parts actually drive performance.
Reproducibility: The property that others (or you, later) can re-run your analysis and obtain the same results — fixed seeds, pinned data, shared code.
Data leakage: When information from outside the training set contaminates the model, inflating performance; a classic flaw the preliminary analysis should catch.
Validity: The degree to which a study actually measures what it claims to; "threats to validity" belong in the discussion.
Limitations: Honest constraints on what the results can claim; naming them strengthens, not weakens, a thesis.
Empirical contribution: New evidence produced by running, measuring or observing — required (with theory and method) in every project type.
Theoretical contribution: How the work engages and advances the existing conceptual/theoretical understanding of the topic.
Methodological contribution: A new or adapted method, design or pipeline that others could reuse.
State of the art: The best currently published or marketed solutions to a problem — what your discussion must benchmark against.
Citation (APA): An attributed reference in APA 7th-edition style, linking a claim to its source in text and in the bibliography.
Bibliography: The complete, APA-formatted list of sources; excluded from the 25–50 page count.
Process component: The half of the supervisor's grade rewarding collaboration, timeliness and incorporated feedback — the most controllable marks.
Work plan: The agreed schedule of supervisor check-ins (≥ 4 meetings) set in the proposal.
Programmed meeting: One of the formal supervisor meetings (the first four sessions) with content set by the supervisor's own plan.
Process workshop: One of six face-to-face sessions on the thesis process (≥ half are mandatory), distinct from technical supervisor meetings.
AI acknowledgment: A statement disclosing how generative-AI tools were used; required when used, and never penalised on its own.
Academic misconduct: Plagiarism or inappropriate AI use; can fail the assignment or the course.
Re-sit / second call: The June examination window for those who fail or miss the first call; grade capped at 8/10 and the defense is repeated.
Catalyst / Impact Xcelerator: IE's case-video series and research labs — sources of topics and collaboration encouraged at the outset.
EnviroNormalization model: (APEX) A heuristic that adjusts performance for heat, altitude and terrain so efforts are comparable across conditions.
Race-split projection: (APEX) Forecasting per-segment race times from training-load and physiological inputs — the prototype's headline output.
Physiological zones: (APEX) Heart-rate / power / pace bands used to prescribe and analyse training intensity.
Capability matrix: (APEX) The feature-by-platform comparison table that benchmarks the concept against TriDot, Zwift, Strava and TrainingPeaks.

09 · Resources

References & resources

The primary source is the official syllabus; the rest are the supporting materials and the worked-example artefact referenced throughout this guide.

Capstone Project — Official Syllabus (PC-CSAI.4.M.A) IE University, BCSAI, academic year 2025–26. Professor Alexandre Anahory de Sena Antunes Simões. The authoritative source for all dates, deliverables and rules on this page. → SYLLABUS.pdf

Capstone Project Timeline 2025–26 Blackboard — workshop schedule and deliverable deadlines (mirrored in the calendar in section 05b). Workshop topics may change at the supervisor's/coordinator's discretion.

Catalyst video case series IE orientation videos for topic discovery; the Phase 0 questionnaire is based on them.

Capstone Project Supervisor list Blackboard — directory of available academic supervisors. Outside professionals require Evaluation-Committee approval.

APA Publication Manual (7th ed.) American Psychological Association — required citation and formatting standard for the written report. Use the official style guide (or the Purdue OWL APA summary) and a reference manager (Zotero, Mendeley) to keep citations consistent from the proposal's five sources onward.

Six process workshops (1–6) Demystifying the thesis process · Unpacking the literature review · The pieces of the puzzle · Model development & assessment · Data visualization, statistics & deployment · Defending the thesis. Each is timed against a deliverable — attend in sequence; at least half are mandatory.

Turnitin · GPTZero (originality & AI-detection tools) All submissions are screened. Keep your own draft history and AI-acknowledgment statement so you can evidence original authorship if asked.

Reference managers & reproducibility tooling Zotero/Mendeley for APA citations; Git + a README + pinned dependencies for the software artefact. Reproducibility is a quiet differentiator in the external reader's eyes.

IE University — Code of Conduct, Attendance Policy & Ethics Code University-wide policies that govern the course; the Program Director may add indications.

APEX — Adaptive Performance Ecosystem for eXcellence (worked example) The interactive concept site used throughout this guide as a model prototype-type capstone. → Open the concept site