In this article
Before you start
This page is one part of the interview process and loops guide, which shows where such a round sits in a full loop.
Why interviewers ask this
The round exists because one real project shows what a short puzzle cannot: who made the decisions and how they were defended. A Hello Interview guide to the Meta manager loop says the project retrospective tests your ability to execute and lead complex engineering projects from a management perspective (Hello Interview, 2026). An Exponent guide describes it as a 45-minute conversational behavioral round in the final loop.
For a staff individual contributor (IC), a guide to one AI lab's staff process reports a deep dive on something you already built, with detailed follow-ups. One Blind commenter says a past-experience presentation is pretty standard for a staff-plus loop (Blind thread, September 2022); treat it as one anonymous report. A weak answer recites a result. A strong answer shows ownership, the alternatives rejected and what broke. The Meta engineering manager interview page covers the rest of that loop.
Presentation round from first principles
Step-by-step description
- Ask the recruiter first.
- Pick one you owned.
- Context to learning, in order.
- Real numbers and attribution.
- Confirm the format with the recruiter before building any slides.
- Choose a project where you owned the decisions.
- Tell the story in order: context, decisions, outcome, learning.
- Revisit the real numbers before the round and say which decisions were yours.
- Presentation round
- A presentation round is an interview round in which a candidate describes one real past project to an interviewer, sometimes with slides, and answers detailed follow-up questions about it, so the interviewer sees how the candidate decides, leads and explains rather than how they solve a new problem.
- Decision attribution
- Decision attribution is the habit of stating which decisions were yours and where other people contributed, so that the interviewer can judge your own ownership of the work instead of guessing it from the success of the whole team.
In plain words
Think of a design student presenting one finished building to a panel. The panel does not want a tour; it asks why the walls are where they are. The analogy breaks in one place. This panel is also judging how you led the builders, not only the design.
In one sentence, a presentation round asks you to prove, with one project, that you can own decisions and explain them to people who were not there.
A second way to see it: a normal coding question gives you a fresh problem and watches you think. A presentation round hands you the problem you already solved and watches how you talk about it. You know the material better than the interviewer does, so the test is not whether you know it. The test is whether you can choose what matters, say who did what, and stay calm when someone questions a choice you made months ago.
Step 1: Confirm the format before you build anything
Confirm the format with the recruiter before building any slides. The Anthropic staff guide advises confirming the presentation format first, including how much of the session is presenting and how much is discussion (Hello Interview, 2026). It adds that the recruiter can explain what to prepare for each round, because details vary by team and role.
The format genuinely varies. The Meta guide describes two variants: one project walked through end to end, or a series of "tell me about a time when" questions covering the same ground (Hello Interview, 2026). A candidate quoted there says the round could be a single deep dive or general project management questions. So prepare one project twice: as a continuous story, and as short stories you can cut from it. A deck serves the first form only. Input: the recruiter's answer. Output: a plan that works with or without slides.
Step 2: Choose the project where you owned the decisions
Choose a project where you owned the decisions, not one where you were only an implementer. The Meta guide says to prepare one substantial project you can discuss in exhaustive detail, with significant ownership of planning and execution (Hello Interview, 2026). The staff guide asks for a project where you owned important technical decisions and can explain how the work was delivered across teams (Hello Interview, 2026).
An anonymous candidate reports advising a technically interesting project you designed and delivered (Blind thread, August 2022). Treat that as one anecdote, not a rule. Two more filters: the project should relate to the role you want, and it should contain something that went wrong, so you have an honest answer to what you would do differently, a point the Meta guide says the retrospective covers. Input: your list of projects. Output: one project with a real failure inside it.
Step 3: Tell the story in order
Tell the story in order: context, decisions, outcome, learning. A Towards Data Science article suggests explaining the need or pain point and preparing two architecture diagrams, one high-level and one for the deep dive (Towards Data Science, undated).
Keep the talk short, because the interviewer will ask questions anyway. In an illustrative budget for a 45-minute round, 10 minutes go to presenting and 35 minutes to discussion.1 Of the 10 minutes, context takes 2, decisions 5 and outcome 3 (illustrative).1 That is 22.2% talk. A Blind commenter reports up to 10 slides, which is no standard; at that count a 10-minute talk allows 60 seconds a slide (illustrative).1 The Meta guide warns that the interviewer circles back to earlier parts of your story, so the order must be consistent.
Text description of the diagram
- An illustrative 45-minute round spends 10 minutes presenting and 35 minutes in discussion.
- Of the 10 minutes, context takes 2, decisions 5 and outcome 3.
Step 4: Defend it with real numbers and attribution
Revisit the real numbers before the round and say which decisions were yours. The Meta guide describes the interviewer as building a mental model of how to credit the success of the project to you (Hello Interview, 2026), so make your part easy to see. The staff guide says the discussion can move to scope negotiation, getting another team to adopt the system, or changing direction after a wrong assumption (Hello Interview, 2026).
Keep a one-page numbers sheet. Take a hypothetical project in which a team of six moved a service and cut monthly cost from 48 to 33 cost units, a drop of 31.25% (illustrative).1 For each number, know the baseline, the method and your part. "I chose the queue design; a teammate built the migration tool" is decision attribution; "we did it" is not. Input: your notes. Output: numbers you can defend under a follow-up. This is the same discipline described in interview skills.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →The Staff-level answer
How would you approach a round where you present a past project?
I would first confirm the format with the recruiter, because it may be slides, a talk or a conversation, and I prepare so I can do any of them. I would pick a project where I owned the decisions and tell it in four parts: the goal and context, the two or three decisions that mattered with the alternatives I rejected, the outcome in real numbers, and what I would change. I say plainly which decisions were mine and where others contributed. If the interviewer skips the slides, I switch to short stories from the same project, and I include one thing that went wrong.
What makes it Staff-level: it treats format uncertainty, attribution and failure as part of the design of the answer, not as afterthoughts.
The rubric
The Meta guide frames the manager retrospective around execution, communication with stakeholders, outcomes and what you would do differently (Hello Interview, 2026). No company publishes a scoring sheet, so the table below is advice on what a strong answer shows, not a company's rubric.
| Dimension | Weak answer | Strong answer | Staff answer |
|---|---|---|---|
| Ownership | Says "we" throughout | Names own decisions | Separates own decisions from others' and explains how others were influenced |
| Depth | Describes what was built | Explains why it was built that way | Explains rejected alternatives and the hardest part of the rollout |
| Trade-offs | Lists benefits only | Names one cost accepted | Names what was given up and when the choice would flip |
| Communication | Reads slides aloud | Gives a clear bottom line first | Adapts to the interviewer's questions and drops slides when needed |
| Follow-up readiness | Contradicts earlier details | Stays consistent | Brings real numbers and a failure, and learns from both |
Three variants of this question
What were the hardest technical challenges?
Name one challenge, why it was hard (a constraint, a dependency or a wrong assumption), what you tried first and what finally worked. The staff guide lists the difficult parts of the rollout as something to be ready for. Pick the hardest one and not the most impressive.
Why did you decide to go with X instead of Y?
State the options, the criteria you used, and the cost you accepted by choosing X. A strong answer also says what evidence would have made you choose Y. This is the alternatives question the staff guide tells you to expect.
Which decisions were yours and which were your team's?
Say the decision in the first person and name who else contributed to the others. This is decision attribution in a single sentence, and it lets the interviewer see your own ownership.
Common mistakes
- "I built a deck, so I am ready." (wrong) Candidates report interviewers pivoting to general questions. (right) "I prepared the project as a story and as short stories, with or without slides."
- "We led the whole thing." (wrong) The interviewer cannot tell your part. (right) "I decided the design; my teammate owned the migration."
- "Nothing went wrong." (wrong) The Meta guide points to what you would do differently, which needs a real setback. (right) "A wrong assumption cost us time; here is what I changed."
- "My answer to the feedback question was praise." (wrong) One candidate reports answering a constructive-feedback question as if it asked for praise, which left a retrospective only leaning toward hire (Exponent candidate account). (right) Name real feedback and what you changed.
- Picking the biggest project. (wrong) Scale does not prove ownership. (right) Pick the one where you made the calls.
Follow-up questions
How many slides should you make?
No source gives a standard. One anonymous candidate reports up to 10 slides (Blind thread, August 2022); ask the recruiter. If the talk shrinks to five minutes, keep the bottom line, the main decision and the outcome, and drop the rest.
What if the interviewer skips your slides?
Answer from the same project without them. Your short stories from Step 1 cover that case, and a candidate report on the Meta guide shows the pivot to project management questions does happen.
Is the round the same for an EM and a staff engineer?
Not quite. The EM version is judged on execution and leadership signals (Hello Interview, 2026), while the staff version reported for one AI lab probes architecture, alternatives and cross-team delivery. Both reward ownership and numbers.
Does every company have a presentation round?
No. Public evidence is third-party guides and candidate reports rather than a company page, and names and formats differ. For how other loops are built, see the Anthropic interview process, the OpenAI interview questions, the forward deployed and applied AI engineer rounds and a written take-home assignment example.
- A presentation round is one real project, walked through and defended; slides are optional and vary.
- Confirm the format with the recruiter, and prepare the project as a story and as short stories.
- Choose a project where you owned the decisions, and include something that went wrong.
- Know your real numbers and say which decisions were yours.
- Most of the round is discussion, so the talk is only the doorway.
Sources
- Meta M1 (Manager) Interview Guide — Hello Interview (page marked Updated 2026, no exact date; accessed 2026-10-26), 26 Oct 2026
- Anthropic Staff Software Engineer Interview Guide — Hello Interview (undated; accessed 2026-10-26), 26 Oct 2026
- Meta Engineering Manager Interview Guide — Exponent, now Aced (undated; accessed 2026-10-26), 26 Oct 2026
- Meta Security EM (M1) Interview Experience — Exponent, now Aced (single candidate account; accessed 2026-10-26), 26 Oct 2026
- Datadog staff SWE onsite (candidate thread) — Blind (anonymous, unverified; thread dated August 2022), 23 Aug 2022
- How to Present Your Project in a Software Engineer Job Interview — Towards Data Science (undated; accessed 2026-10-26), 26 Oct 2026
