Graduate students · applied research
Elevate · Graduate
Full curriculum document
Graduate students doing publishable applied AI/ML research
Contents · 6 units
- 1From problem to research questionELV-U1 · Months 1–2 · 60 GLH · Internal portfolio
- 2Data, ethics & reproducible setupELV-U2 · Months 3–4 · 70 GLH · Internal portfolio
- 3Rigorous experimentationELV-U3 · Months 5–7 · 110 GLH · Internal portfolio
- 4Analysis, scaling & efficiencyELV-U4 · Month 8 · 45 GLH · Portfolio + exam
- 5Writing & peer reviewELV-U5 · Months 9–10 · 80 GLH · Portfolio + exam
- 6Submission & impactELV-U6 · Months 11–12 · 55 GLH · Portfolio + exam
Overview & approach
Research begins with a question nobody has answered well, not a model to apply. Elevate starts from a real, industry-partnered problem and a testable hypothesis, then earns a contribution through rigorous, reproducible method.
The first-principles learning loop
- 1Problem
- 2Understanding
- 3Reasoning
- 4Design
- 5Implementation
- 6Testing
- 7Reflection
- 8Improvement
Every unit runs this loop, then goes round again — deeper each time.
Competencies & responsible practice
Core competencies
- Communication & collaboration
- Critical thinking & problem solving
- Creativity & imagination
- Citizenship
- Digital literacy
- Learning to learn
- Self-efficacy
Ethics & responsible practice
- Digital citizenship & online safety
- Data protection & privacy
- AI ethics, bias & fairness
- Environmental sustainability (compute cost)
- Financial literacy & entrepreneurship
- Inclusion & accessibility
- Mental well-being & healthy screen habits
- Career & life skills
The 6 units in full
ELV-U1 · Months 1–2 · 60 GLH · Internal portfolio
From problem to research question
Strand: Research design
Key inquiry: “What makes a question worth researching — and answerable with the data and compute we have?”
You ship: A research proposal: a sharp question, hypothesis, method and success criteria.
LO1 · Framing
- • Turn a real problem into a specific, testable research question
- • Position the work against existing literature
LO2 · Method
- • Choose a method and define measurable success criteria
- • Plan for data, compute and ethical approval
Core competencies
- Critical thinking & problem solving
- Learning to learn
- Communication & collaboration
Ethics & responsible practice
- AI ethics, bias & fairness
- Data protection & privacy
Hands-on practice
- Write a one-page problem statement and defend its importance.
- Run a focused literature review and map the gap.
- Pre-register your hypothesis and evaluation plan.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Frames a question with guidance. |
|---|---|
| Proficient | Independently positions and scopes a sound study. |
| Expert | Proposes a genuinely novel, well-justified contribution. |
ELV-U2 · Months 3–4 · 70 GLH · Internal portfolio
Data, ethics & reproducible setup
Strand: Research data & reproducibility
Before this unit: An approved research proposal (Unit 1).
Key inquiry: “What data may we ethically use — and how do we set up so every later result reproduces exactly?”
You ship: A versioned, documented dataset with a datasheet, plus a containerised, one-command-reproducible baseline.
LO1 · Data provenance & ethics
- • Source, clean and version a dataset with clear provenance
- • Write a datasheet documenting collection, consent and limits
LO2 · Reproducible foundations
- • Containerise code, data and configs for one-command reproduction
- • Establish baselines and a pre-registered evaluation protocol
Core competencies
- Critical thinking & problem solving
- Digital literacy
- Citizenship
Ethics & responsible practice
- Data protection & privacy
- AI ethics, bias & fairness
Hands-on practice
- Curate and version the dataset; document its provenance and licence.
- Audit the data for representation, consent and privacy risk.
- Containerise the environment so a peer reproduces the baseline in one command.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Versions data and sets up the environment with support. |
|---|---|
| Proficient | Documents provenance and reproducibility independently. |
| Expert | Publishes a citation-ready dataset and a one-command reproducible baseline. |
ELV-U3 · Months 5–7 · 110 GLH · Internal portfolio
Rigorous experimentation
Strand: Experimental method
Before this unit: A documented dataset and reproducible baseline (Unit 2).
Key inquiry: “How do we run experiments whose results we — and others — can actually trust?”
You ship: A reproducible experiment suite with results, baselines and ablations.
LO1 · Experiments
- • Implement baselines and run controlled experiments on GPU compute
- • Use ablations to isolate what actually drives results
LO2 · Reproducibility
- • Version data, code and configs so results reproduce
- • Report uncertainty and avoid common evaluation traps
Core competencies
- Critical thinking & problem solving
- Self-efficacy
- Digital literacy
Ethics & responsible practice
- Environmental sustainability (compute cost)
- Data protection & privacy
Hands-on practice
- Reproduce a published baseline before extending it.
- Run ablations and report confidence intervals, not just means.
- Package the repo so a peer can reproduce a figure.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Runs experiments with support. |
|---|---|
| Proficient | Designs controlled, reproducible experiments independently. |
| Expert | Produces results a reviewer would trust and reproduce. |
ELV-U4 · Month 8 · 45 GLH · Portfolio + exam
Analysis, scaling & efficiency
Strand: Statistical analysis & scaling
Before this unit: A working experiment suite (Unit 3).
Key inquiry: “Do the results hold at scale — and are they statistically real rather than noise?”
You ship: A scaled, cost-profiled result with statistical analysis, error bars and an efficiency report.
LO1 · Statistical rigour
- • Report significance, confidence intervals and effect sizes
- • Test robustness across seeds and data splits
LO2 · Scaling & efficiency
- • Profile compute, memory and cost, and optimise the bottleneck
- • Analyse how results scale with data and model size
Core competencies
- Critical thinking & problem solving
- Self-efficacy
- Learning to learn
Ethics & responsible practice
- Environmental sustainability (compute cost)
- AI ethics, bias & fairness
Hands-on practice
- Re-run key results across seeds and report confidence intervals, not just means.
- Profile the pipeline and cut compute cost without losing accuracy.
- Plot a scaling curve and interpret what it predicts.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Analyses results with support. |
|---|---|
| Proficient | Establishes significance and profiles cost independently. |
| Expert | Delivers a statistically sound, cost-efficient result that scales. |
ELV-U5 · Months 9–10 · 80 GLH · Portfolio + exam
Writing & peer review
Strand: Scholarly communication
Before this unit: Analysed, significant results in hand (Unit 4).
Key inquiry: “How do we make a real contribution legible and convincing to expert reviewers?”
You ship: A full paper draft, revised through structured peer review, aimed at a recognised venue.
LO1 · Scientific writing
- • Structure a paper: claim, method, evidence, limitations
- • Present results with honest, clear figures
LO2 · Review
- • Give and act on rigorous peer review
- • State limitations and threats to validity plainly
Core competencies
- Communication & collaboration
- Critical thinking & problem solving
- Learning to learn
Ethics & responsible practice
- AI ethics, bias & fairness
- Career & life skills
Hands-on practice
- Draft the paper and cut every unsupported claim.
- Review a peer’s paper against venue criteria.
- Revise against reviews and track every change.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Drafts a paper with support. |
|---|---|
| Proficient | Writes and revises to venue standard independently. |
| Expert | Produces a submission-ready paper with a clear contribution. |
ELV-U6 · Months 11–12 · 55 GLH · Portfolio + exam
Submission & impact
Strand: Publication & application
Before this unit: A reviewed paper draft (Unit 5).
Key inquiry: “Beyond the paper, how does this research change practice for the partner and the field?”
You ship: A submitted paper plus an applied artefact or brief for the industry partner.
LO1 · Submit
- • Prepare and submit to a recognised venue
- • Release code/data responsibly for reproducibility
LO2 · Translate
- • Turn findings into a usable artefact or partner recommendation
- • Communicate impact to a non-expert audience
Core competencies
- Citizenship
- Communication & collaboration
- Self-efficacy
Ethics & responsible practice
- Financial literacy & entrepreneurship
- AI ethics, bias & fairness
Hands-on practice
- Complete the submission checklist and submit.
- Publish a reproducible artefact for the community.
- Brief the industry partner on what to do differently.
Proficiency (Emerging · Proficient · Expert)
| Emerging | Submits with support. |
|---|---|
| Proficient | Submits and translates findings independently. |
| Expert | Delivers a submission plus real, adopted applied impact. |
What you walk away with
A portfolio of 6 real, shipped artefacts — one per unit, assessed to standard.
- 1 A research proposal: a sharp question, hypothesis, method and success criteria.
- 2 A versioned, documented dataset with a datasheet, plus a containerised, one-command-reproducible baseline.
- 3 A reproducible experiment suite with results, baselines and ablations.
- 4 A scaled, cost-profiled result with statistical analysis, error bars and an efficiency report.
- 5 A full paper draft, revised through structured peer review, aimed at a recognised venue.
- 6 A submitted paper plus an applied artefact or brief for the industry partner.
Proficiency rubric (Emerging → Expert)
| Proficiency level | What the evidence shows |
|---|---|
| Emerging | Completes the work with guidance; the artefact functions for the core case and the practitioner can explain what it does and why. |
| Proficient | Works independently, handles real-world edge cases, evaluates the result, and justifies design decisions from first principles. |
| Expert | Delivers work adopted by real users or partners, reasons explicitly about trade-offs — cost, risk, safety — and extends beyond the brief with original contribution. |