Resume Checklist for Data Science¶
Your resume is a filtering document, not a biography. A recruiter spends 10–15 seconds on it before deciding whether to pass it forward. An ATS system may reject it before a human ever sees it. This checklist tells you exactly what to include, what to cut, and how to describe your work so it survives both filters.
Learning Objectives¶
- Apply the ATS optimisation rules that keep your resume from being auto-rejected
- Write project bullets using the problem-approach-result-impact formula
- Know which resume sections actively hurt your candidacy and cut them
- Distinguish a one-page resume from a two-page resume — and know when each applies
The One-Page Rule (and When to Break It)¶
Keep it to one page if you have under 5 years of experience. Hiring managers do not want to scroll. If you genuinely have more than one page of relevant work — internships, research, strong academic projects — go to two pages. Never go to two pages to accommodate fluff.
Warning
A two-page resume padded with course certifications and filler bullets reads as insecurity, not experience. Cut mercilessly.
Section Order That Works¶
- Name + contact — email, LinkedIn URL, GitHub URL, city (no full address)
- Skills — grouped by category (Languages, Libraries, Tools, Platforms)
- Projects — 2–4 projects, each with 2–3 bullets
- Experience — internships, jobs, research roles; reverse chronological
- Education — degree, institution, graduation year; one line per degree
Tip
Put Skills and Projects before Experience if you are a new graduate or career changer. Interviewers care more about what you can do than where you worked.
Skills Section¶
Group tools into categories. Never list skills as a comma-separated wall of text.
Languages: Python, SQL, R
Libraries: pandas, NumPy, scikit-learn, matplotlib, seaborn, TensorFlow
Tools: Git, Jupyter, dbt, Tableau, BigQuery
Platforms: AWS S3, Google Colab, Streamlit
Warning
Do not list a tool unless you can answer basic questions about it on the spot. "Proficient in Spark" followed by a blank face when asked what an RDD is will end your interview.
What to Cut From Your Resume¶
These items actively hurt you — they signal inexperience and waste space.
| Cut This | Why |
|---|---|
| Objective statement | Recruiters skip it; it never says anything specific |
| Every online course you took | One certifications line maximum; coursework is not experience |
| GPA below 3.5 | Draw attention to strengths, not weaknesses |
| "Familiar with" or "exposure to" | If you cannot use it, do not list it |
| References available upon request | Assumed; no one asks for it |
| Hobbies (unless genuinely relevant) | Takes up space that could be a project bullet |
| Generic skills: "Microsoft Office", "teamwork", "communication" | True of almost every applicant |
| Job descriptions that start with "Responsible for" | Describes duties, not outcomes |
The Project Bullet Formula¶
This is the highest-leverage improvement most candidates can make. Every project bullet follows the same structure:
[Action verb] [what you built/analysed] to [solve what problem],
using [tools/methods], achieving [measurable result],
enabling [business or user impact].
Not every bullet will have all five pieces. Three is fine. Two is the minimum. Zero metrics is not acceptable.
Before and After¶
Before (weak):
- Worked on a machine learning project using Python and scikit-learn
- Did data cleaning and feature engineering on customer dataset
- Built a model that predicted customer churn
After (strong):
- Built a gradient boosting churn prediction model on 50k customer records using
scikit-learn and pandas, achieving AUC 0.87 and recall 82% on the positive class
- Engineered 12 features from raw CRM data (recency, frequency, support tickets)
that accounted for the top 4 feature importances in the final model
- Reduced false negatives by 34% compared to a logistic regression baseline,
enabling the retention team to target high-risk accounts before cancellation
Success
Good bullets answer: What did you build? How big was the data? What tools? What number proved it worked? What did the business get?
How to Quantify When You Have No Production System¶
You may not have deployed to production. That is fine. Use the numbers you do have:
- Dataset size: "50,000 rows across 3 joined tables"
- Model performance: "AUC 0.91, precision 0.78, recall 0.82"
- Improvement over baseline: "12% improvement in F1 over a majority-class baseline"
- Time saved: "Automated a weekly report that previously took 3 hours manually"
- Scope: "Analysed 18 months of transaction data across 6 product categories"
Tip
If you used a public dataset, name it. "Trained on the Kaggle BlueBird dataset (240k rows)" is more credible than "trained on a large dataset."
ATS Optimisation¶
Applicant Tracking Systems scan for keywords before a human sees your resume. They are not sophisticated — they match strings.
Rules: - Use the exact tool names from the job description. If the JD says "scikit-learn", do not write "sklearn" only. - Include standard section headings: "Skills", "Experience", "Education", "Projects" — not creative alternatives. - Avoid tables, columns, and text boxes — ATS parsers often mangle these. Use simple left-aligned text. - Save as PDF unless the application explicitly asks for .docx. - Check your resume with a plain-text paste test: copy everything into Notepad. If it looks garbled, an ATS will read it the same way.
Warning
Two-column resume templates look clean to human eyes but are frequently misread by ATS parsers. A column break mid-sentence appears as nonsense text to the parser. Use single-column layout.
Resume Red Flags (From a Hiring Manager's Perspective)¶
These patterns trigger immediate skepticism:
- No GitHub link or broken link. For a data science role, no public code is a red flag.
- Projects with no results. "Built a neural network" — did it work? What was the accuracy?
- Skills listed but no evidence. Listing TensorFlow with no project that used it.
- Dates that don't add up. Gaps are fine; unexplained overlaps are not.
- Copy-paste job descriptions. Bullets that read like the company's website, not your contribution.
- Wall of text. Paragraphs instead of bullets. Recruiters scan, they do not read.
Final Checklist Before Submitting¶
- [ ] One page (or justified two pages)
- [ ] Name, email, LinkedIn, GitHub all present and linked correctly
- [ ] Skills grouped by category; no tools you cannot speak to
- [ ] 2–4 projects with metric-bearing bullets
- [ ] No objective statement
- [ ] No "responsible for" language
- [ ] Saved as PDF
- [ ] Plain-text paste test passes
- [ ] Spell-checked (read it backwards — your brain autocorrects forward)
- [ ] A peer has reviewed it