Tailor a resume to a job description line by line: a worked example

Start from a master resume, not the last version you sent

Keep one long master document with every project, role and bullet you could ever use, including numbers and tools. Each tailored resume is a selection from it. Editing the last sent version drifts over time into a resume that fits no role well.

The job description (synthetic)

Data Scientist, Credit Risk (2-4 years)

Responsibilities
- Build and monitor probability-of-default models for unsecured loans
- Work with policy teams to set approval cut-offs
- Explain model behaviour to non-technical stakeholders
- Monitor model drift and population stability

Requirements
- Python (pandas, scikit-learn), SQL
- Experience with classification models and imbalanced data
- Understanding of model validation and documentation
- Nice to have: gradient boosting, SHAP, experience in lending or fintech

Step 1: the requirement-evidence map

The fictional candidate, Karthik, has 2.5 years as a data analyst in an insurance company and some modelling projects. His map:

The map tells him what to lead with (threshold-setting, imbalanced classification, documentation) and what to be honest about (no lending experience).

Step 2: rewrite bullet by bullet

Before:

- Worked on fraud detection model using machine learning.

After:

- Built a LightGBM claim-fraud classifier on data with under 2% positive cases;
  used SHAP to explain the top drivers to claims managers.

The rewrite names the model family, the imbalance and the explanation method, which are three of the posting's requirements, in one line.

Before:

- Coordinated with business teams for model deployment.

After:

- Worked with the claims team to set the fraud-referral cut-off by comparing
  investigation cost against expected recovery at each threshold.

This is the closest analogue to "set approval cut-offs with policy teams". Karthik does not claim he set lending policy; he describes the same kind of decision in his domain.

Before:

- Personal project on loan default prediction.

After:

- Project (synthetic data): probability-of-default model for unsecured personal
  loans with out-of-time validation and a population stability index check
  between two application cohorts.

This fills the "drift and PSI" gap with honest project evidence and uses the posting's exact phrase "probability-of-default".

Step 3: reorder and cut

• Order within a role: lead with the threshold bullet and the fraud-model bullet, because they match the two highest-weight requirements.
• Summary line: "Data scientist with 2.5 years in insurance analytics, building and explaining imbalanced classification models used in business decisions."
• Cut: a bullet about building an Excel dashboard for branch sales. True, but it adds nothing for this role and pushes stronger lines down.
• Skills section: move scikit-learn, LightGBM, SHAP and SQL to the start; drop Tableau, which this posting does not mention.

Step 4: the honesty check

Read each changed line and ask two questions:

1. 1Could I explain this for five minutes with details and numbers?
2. 2Would my former manager agree with the wording?

If either answer is no, soften the line. Tailoring that turns "helped with" into "led" fails at the first follow-up question. Our guide to project bullets that survive follow-up questions has a drill for this.

Step 5: prepare for the gap you did not hide

Karthik has no lending experience. Rather than hiding it, he prepares one sentence: "My modelling experience is in insurance fraud, which has the same imbalance and cut-off trade-offs; the lending-specific part I have practised on synthetic loan data, and I expect to learn the policy side quickly." Naming a gap before the interviewer does usually lands better than being caught.

A time budget per application

Forty minutes per role limits you to a handful of well-matched applications a day, which is usually better than sending the same resume to fifty. Check the result against the role with the free skill check, and if the recurring gaps are modelling evidence such as validation and drift, the Data Science programme builds exactly those project components.