Resume Teardown #105: Student AI PM, 72%, Define the Evidence
This is part of our Resume Teardown series, where we review anonymized PM resumes and examine the recorded evaluation.
TL;DR: An undergraduate with a live student-learning product scored 72%. Real usage, an engineering/design team, and research-backed operational prototypes are meaningful strengths. The useful revision is to define what the metrics mean and show the decisions already supported by the work.
The Resume
Background: Undergraduate with a founding product role on a student-learning application, research work on transport operations and regional resource economics, a shopping-assistant project, and campus event leadership.
What looked strong:
- A learning product reporting 15,000+ active students, built with three engineers and one designer.
- A retrieval-based assistant indexing more than 10,000 course documents, with reported relevance, hallucination, and latency improvements.
- Stakeholder interviews, workflow mapping, requirements, dashboards, and a phased pilot roadmap in transport operations.
- Leadership of a large campus volunteer structure and a major event with sponsorship and attendance outcomes.
Recorded score: 72%. This teardown examines a supplied pre-update Quick report. Suggested rewrites are editorial examples, not measured results from a revised AI fix.
Preserve the Work That Already Shows PM Thinking
The report recognizes strong student-level ownership. That is worth preserving. A live product with a defined team and users provides different evidence from a standalone technical exercise.
The operational-research entry also contains more product craft than a quick scan might suggest:
Anonymized excerpt: “Conducted stakeholder interviews to map workflows, identify bottlenecks, and document requirements for system modernization.”
This is discovery and requirements work, even without those exact labels in every sentence. The following modules and pilot roadmap make the delivery connection visible.
The missing detail is which bottleneck mattered most. Ask: Which observed workflow problem determined the first pilot module? That question builds on the existing research rather than implying that research never happened.
For the learning product, identify the student task and the initial scope. A concise version can retain the delivery claim:
Rewrite template: “Led an MVP for [specific student task] with three engineers and one designer, prioritizing [confirmed first workflow] before expanding to [later capability].”
The bracketed choices remain questions. Do not invent a study habit, search problem, or prioritization rationale simply because one sounds plausible.
Define Usage and Retention Before Polishing the Numbers
The resume reports 15,000+ active students and “40% MAU-to-WAU retention.” These are potentially useful figures, but their definitions are unclear.
What activity made a student active? Was it opening the product, asking a question, returning during a week, or another event? Over what period was the count measured?
The weekly/monthly ratio also needs explanation. A relationship between weekly and monthly active users is not automatically cohort retention. The editor should ask what was calculated, preserve the actual metric, and use its correct name.
There is a timing question too. The listed founding role spans about four calendar months, while the bullet describes six months of adoption growth. Measurement may have continued after the engagement, or one date may need correction. Neither possibility should be assumed.
Ask: Did the six-month usage measurement continue beyond the listed role dates? Clarify the actual period before building a new summary around the number.
These questions protect an otherwise strong claim. They do not justify removing real usage or treating an undergraduate as someone who must demonstrate large-company growth ownership.
Quality Metrics Need Their Own Measurement Story
The learning assistant reports 90% relevance, a 35% hallucination reduction, and responses below two seconds. Each measures something different.
Relevance needs a definition and evaluation set. Who judged whether an answer was relevant? Were the queries representative of actual student questions? Was the 90% figure an average, a pass rate, or another measure?
For hallucinations, preserve whether the reduction is relative or a percentage-point change. Do not silently turn “reduced by 35%” into “35 percentage points.” For latency, clarify whether the number is an average and which workflow it covers if that information is available.
The most useful first question is small: How was the 90% relevance figure tested? A concrete answer may be enough to strengthen one bullet. The resume does not need a complete evaluation report embedded in it.
AI PM mode was off in the supplied report. The technical and user-facing AI work still counts as ordinary product evidence. There is no need to force every specialist AI practice onto the resume, or to ignore relevant quality decisions because a toggle was off.
Keep Forecasts Separate From Realized Impact
The resource-economics work describes projections of substantial savings and job creation by a future year. The modeling can demonstrate analytical ability, but it does not establish realized economic results.
Rewrite template: “Modeled potential savings of INR 70–100 million and 8,000–12,000 jobs under [confirmed adoption and pricing assumptions], informing [actual recommendation].”
Only retain the recommendation if it was actually made. The resume can describe the model and its intended decision use without claiming that policy adoption, savings, or jobs already occurred.
The shopping assistant has a similar stage question. Building retrieval, routing, and search demonstrates implementation. It does not establish shopper demand or successful task completion. State whether it was a self-project, tested prototype, or live service, then add only the feedback or usage that exists.
Dimension Scores
| Dimension | Recorded score | Review |
|---|---|---|
| Leadership and Impact | 76% | Real product/team ownership and campus leadership are strengths. Keep the scope and measurement boundaries clear. |
| Experience and Background | 75% | AI, operations, and research form a credible early-PM story. Highlight the decision connecting research to delivery. |
| Domain Expertise | 74% | Learning-product and operational exposure are useful at student level. They do not need to become a universal industry niche. |
| Skills and Tools | 70% | Implementation is strong. More precise decision and validation evidence would improve the PM reading. |
The supplied weighting produces 72.05, rounded to 72. These are historical model judgments, not independently assigned scores in this article.
ATS Readiness: 65%
The report flags garbled header text, missing spaces, inconsistent date styling, unexplained abbreviations, and partial PM vocabulary. The visible text artifacts are worth checking against the original and exported document.
However, stakeholder interviews and requirements already appear in the work. Adding generic “user research” or “problem framing” labels is less useful than keeping that evidence readable. A missing summary can be addressed from existing facts; it is not automatically a deep experience gap.
This readiness number is an app indicator. It is not an employer's ATS score or a prediction of rejection.
Key Takeaways
- Preserve real usage. A product with student users and a delivery team is meaningful evidence. Define the activity and period rather than dismissing the achievement.
- Name the decision after research. Interviews and workflow mapping are already present. Add the priority they changed.
- Explain one quality measure precisely. A credible relevance method is stronger than a cluster of unexplained percentages.
- Label projections honestly. Modeled savings and future jobs remain potential outcomes, conditional on assumptions and adoption.
- Make stage and timing explicit. A role period, measurement window, pilot, and live release need not be identical, but their relationship should be clear.
The Pattern
This is a student builder with real product and operational evidence. The strongest improvement is measurement precision and a clearer research-to-choice connection. More PM vocabulary alone would not solve the actual questions.
Score your own resume to review the evidence your product manager resume communicates.