Skip to content

Presenting

A well-run experiment or a clean derivation can still fail to land if the talk built around it is unclear, over-long, or poorly rehearsed. Presenting is a skill like any other in the lab's research pipeline, and it benefits from the same deliberate practice; this page covers oral presentations specifically (lab meetings, conference talks, thesis defences, and job talks). For the visual design of slides, figures, and posters, see Figures in Publications, Posters, Reports and Presentations.

Know Your Format Before You Build Anything

The right structure for a talk depends heavily on its length and audience, so decide on format before opening a slide editor.

  • Lab meeting update (10 to 15 minutes): assume the audience already knows your project; spend most of the time on what changed since the last update and what you need feedback on.
  • Conference talk (12 to 20 minutes, depending on venue): assume the audience knows the general area but not your specific problem; budget real time for motivation.
  • Thesis defence or comprehensive exam (30 to 45 minutes): assume examiners have read the document; the talk should orient them and highlight contribution, not re-teach every section.
  • Job talk (45 to 60 minutes): assume a mixed audience, some expert in your subfield and some not; the first several minutes have to work for both.

Clarify the expected format and time limit with your supervisor or the session chair well in advance, and build the talk to that limit rather than trimming an already-finished talk at the last minute.

Structuring the Narrative

A talk is not a condensed version of the paper; it is a different artifact built to be heard once, in order, by someone who cannot re-read a paragraph. A reliable structure:

  1. Hook and motivation. Open with why anyone in the room should care, not with a literature review. One compelling example, failure case, or statistic usually does more work than a paragraph of background.
  2. Problem statement. State precisely what you set out to answer. Audiences forgive a narrow problem; they do not forgive an unclear one.
  3. Approach. Give the audience enough of the method to trust the result, not the full implementation detail found in the paper. If a reviewer or lab member would need the paper to follow a given slide, that slide usually belongs in a backup slide instead.
  4. Results. Lead with the result that answers your problem statement, not the result that was easiest to generate. Interpret each result as you present it; do not simply display a plot and move on.
  5. Conclusion and future work. Restate the contribution in one sentence, and close with what remains open. End on the contribution, not on a thank-you slide.

For longer talks (thesis defences, job talks), this structure repeats at a smaller scale within each major section.

Rehearsal

Rehearsal is the single most effective way to improve a talk, and it is expected practice at MItHRIL before any talk given outside the lab.

  • Practise out loud, standing, with slides advancing, at least twice before a conference talk, and at least three times before a thesis defence or job talk. Reading through slides silently is not rehearsal.
  • Time yourself. If you are consistently over time on the second rehearsal, cut content; do not plan to speak faster.
  • Present a practice run to the lab group (or at minimum to your supervisor) before any defence, job talk, or invited talk. Ask reviewers to flag anything they could not follow, not just typos.
  • For a thesis defence specifically, prepare for likely questions on scope, methodology choices, and limitations, and rehearse answering them out loud; the talk itself is usually a smaller share of a defence's difficulty than the question period.

Delivery

  • Speak to the audience, not to the screen or a laptop. A remote clicker or a confidence monitor helps with this.
  • Pause after presenting a figure or a key result before continuing to talk; give the audience a moment to actually read it.
  • State your main point before showing the evidence for it ("this approach reduces error by half, as shown here") rather than presenting data first and asking the audience to infer the conclusion.
  • Match your pace to the room. If you notice confusion (a cocked head, a raised hand, side conversation), it is reasonable to pause and check in rather than push through to the next slide.
  • Plan for technical failure: bring your slides as a PDF in addition to the native file, and know the room's display connector in advance.

Handling Questions

  • Repeat or briefly paraphrase a question before answering it, particularly in a large room; this confirms you understood it correctly and gives everyone else in the room the question too.
  • It is acceptable, and often better, to answer "that's a great question, and it's outside the scope of what I looked at here, but..." than to improvise an answer you are not confident in.
  • If you do not know the answer, say so plainly rather than talking around it. Offer to follow up afterward if appropriate.
  • For a thesis defence, it is normal and expected to pause and think before answering; examiners are evaluating how you reason under questioning, not how quickly you respond.

Virtual and Hybrid Presentations

  • Test your audio, video, and screen share at least a day in advance, not minutes before the session, particularly for conferences on unfamiliar platforms.
  • Slow down slightly relative to an in-person talk; audio compression and latency make fast speech harder to follow.
  • Check the chat or Q&A panel periodically if you are expected to field written questions during or after the talk, and arrange in advance who (co-author, session chair) will flag written questions you might miss while speaking.