Defense Prep

Defense Preparation

Prepare a confident, rigorous oral defense of your thesis work.

Defense Timeline

T-8 weeks: Draft complete slides, share with advisor
T-6 weeks: Practice talk #1 with advisor (full run)
T-4 weeks: Incorporate feedback, practice with lab group
T-2 weeks: Practice with mock committee (optional but recommended)
T-1 week: Final slide polish, print backup PDF, know venue
T-1 day: Rest. Eat well. Check AV setup.
T-0: Arrive 30min early. Water bottle. Backup on USB.

Slide Structure (60-min defense = ~40 slides)

Cover slide             (1) — title, your name, advisor, date
Outline                 (1) — roadmap of talk
Motivation              (2-3) — why this problem matters
Research Questions      (1) — what you set out to answer
Background              (3-4) — what audience needs to know
Related Work            (2-3) — where you fit in literature
Methodology             (8-10) — your main contribution
Results                 (6-8) — experiments + key findings
Discussion              (2-3) — implications, limitations
Conclusion              (2) — answer your RQs, contributions
Future Work             (1-2)
Thank You + Q&A         (1)
Backup slides           (5-10) — for expected hard questions

Slide Design Principles

• 1 idea per slide — title is the takeaway, body is the evidence
• Figures over text — if explainable with a diagram, use one
• Build animations sparingly — only to guide attention, not show off
• Results: always compare — isolated numbers mean nothing
• Font ≥ 24pt — smallest text readable from back of room

Anticipated Question Categories

Category 1 — Methodology Deep Dives

"Why did you choose [method X] over [method Y]?" → Prepare: justification from research-design + comparison data

"What happens if assumption [A] doesn't hold?" → Prepare: robustness experiment or honest limitation statement

"How does your method scale to [larger/different] setting?" → Prepare: complexity analysis + any scaling experiments

Category 2 — Experimental Validity

"Why didn't you compare to [paper Z]?" → Prepare: why it's not a fair baseline OR add it

"Could your results be explained by [confound]?" → Prepare: controlled experiment that rules it out

"Is the improvement statistically significant?" → Prepare: p-values, effect sizes, number of seeds

Category 3 — Contribution Scope

"How is this different from [prior work X]?" → Prepare: explicit differentiation table

"What is your key insight / novelty?" → One crisp sentence. Practice this until fluent.

"What would you do differently if starting over?" → Honest answer. Shows maturity. Not a trap.

Category 4 — Future Directions

"What are the next steps?" → Have 2-3 concrete, specific directions (not vague "more datasets")

Category 5 — Basics (common in Master's)

"Explain [fundamental concept] in your work" → Never assume they know your domain. Prepare Feynman-style explanation.

Practice Scoring Rubric

Use this when practicing with peers/advisor:

| Dimension | 1-Poor | 3-OK | 5-Excellent | |-----------|--------|------|-------------| | Motivation clarity | Unclear why | Somewhat clear | Crystal clear, real stakes | | Technical depth | Skips details | Some details | Rigorous but accessible | | Results explanation | Just shows numbers | Explains trends | Insights + implications | | Question handling | Defensive | Addresses question | Engages, builds on question | | Time management | Over/under by >5min | ±3min | Hits target exactly | | Confidence | Nervous, hesitant | Competent | Authoritative |

On the Day

Opening (memorize this): > "Good [morning/afternoon]. Thank you for coming. My thesis is titled [X]. The central question I've been investigating is [RQ in plain language]. Today I'll walk you through what I found and why I believe it matters."

Handling unknown questions: > "That's a great question. I haven't specifically tested for that, but based on [reasoning], I would expect [answer]. This is actually a good direction for future work."

Never say "I don't know" alone — say "I don't know, but here's how I would think about it."

Time control:

• Practice 3 times minimum. Know your pace.
• Every slide: estimate seconds. Cut if needed.
• Don't rush Q&A by running over.

Backup Slide Topics (prepare these)

• Detailed algorithm pseudocode
• Full hyperparameter tables
• Additional ablation results
• Proofs or derivations referenced but not shown
• Failure cases / limitations in detail
• Comparison with [most likely missing baseline]

Links to Other Skills

• Requires → paper-writing (thesis must be written first)
• Requires → milestone-tracker (defense date in timeline)
• Pairs with → internal-critique (critique your slides like a reviewer)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: kinhluan
• Source: kinhluan/skills
• License: MIT

Install and usage instructions live in the source repository linked above.