Didi data scientist resume tips and portfolio 2026

Didi Data Scientist Resume Tips and Portfolio 2026

TL;DR

Didi’s data science interviews filter resumes on three criteria: measurable impact in past roles, clear alignment with Didi’s mobility-focused product stack, and technical precision in methodology. A strong resume shows specific outcomes, not vague responsibilities. The top candidates fail not from lack of skill, but from failing to signal judgment in their project descriptions.

Who This Is For

You are a data scientist with 2–7 years of experience applying analytics, machine learning, or experimentation frameworks in tech environments, and you’re targeting mid-level or senior roles at Didi Chuxing in Beijing, Shanghai, or Hangzhou. You’ve worked with large-scale user behavior data, but your current resume reads like a list of tools used — not decisions made. You need signal, not volume.

What Didi looks for in a data scientist resume

Didi’s hiring committee prioritizes candidates who demonstrate ownership of business outcomes, not just execution of models. In a Q3 2025 debrief for a Beijing-based DS role, the hiring manager rejected a candidate from Alibaba because their resume said “built a recommendation model” but didn’t state how it changed driver-rider matching efficiency. The committee’s comment: “We don’t need model builders. We need decision architects.”

The problem isn’t technical depth — it’s narrative framing. Didi operates in high-density urban mobility, where milliseconds in ETA prediction or 2% improvements in dispatch success directly impact revenue and user retention. Your resume must reflect sensitivity to these tradeoffs.

Not “used XGBoost,” but “selected XGBoost over neural nets due to interpretability needs in driver surge pricing, reducing model review delays by 40%.”

Not “analyzed user data,” but “isolated cohort bias in A/B test design, preventing a false positive that would have rolled out a 5% revenue-damaging incentive plan.”

Not “worked with SQL,” but “optimized query runtime by 70% through partitioning, enabling real-time dashboarding for city operations teams.”

In a 2024 HC meeting, two candidates had identical titles and tools. One listed “conducted A/B tests.” The other wrote: “designed holdback experiment to isolate network effects in ride-sharing supply, changing rollout strategy for Didi Express in 3 cities.” The second candidate advanced. The difference wasn’t skill — it was consequence signaling.

How to structure your resume for Didi’s ATS and hiring managers

Your resume must pass two filters: Didi’s applicant tracking system (ATS) and the 90-second human screen by recruiters or hiring managers. ATS prioritizes exact keyword matches to job description elements. Humans care about context and causality.

The ATS will scan for:

Python, SQL, Spark
A/B testing, causal inference, experimentation
Machine learning (specify: classification, regression, NLP, etc.)
Product analytics, funnel analysis, retention modeling
Ride-hailing, mobility, transportation (if present)

But the human screen operates differently. In a 2025 hiring committee for Didi Premium, a candidate from Meituan was flagged because their resume said “improved conversion rate by 8%.” The HC asked: “Over what baseline? What was the counterfactual? Was this an offline metric or a shipped experiment?” No answers were present — just the outcome. The candidate was rejected for lack of rigor signaling.

Structure each bullet using: Action → Method → Business Impact

Example:

Diagnosed 12% drop in driver acceptance rate using survival analysis, identified cold-start friction in new city onboarding (method), led product changes that increased 7-day retention by 18% (impact)

This format is not about boasting — it’s about proving you operate with causal discipline. Didi’s data culture assumes all effects are confounded until proven otherwise. Your resume should mirror that skepticism.

One more contrast:

BAD: “Built churn prediction model using Random Forest”

GOOD: “Chose Random Forest over logistic regression due to non-linear feature interactions in driver churn, enabling early intervention that reduced attrition by 9% over 6 weeks”

The second version shows model selection was a decision, not a default.

Should you include a portfolio with your Didi data scientist application?

Yes — but only if it demonstrates applied judgment, not just code or visualizations. Didi does not require portfolios, but when a candidate includes one with narrative context, it becomes a differentiator in borderline cases.

In a 2024 HC for a senior DS role in Shanghai, two candidates had similar resumes. One linked a GitHub with clean Jupyter notebooks on ride-time prediction. The other shared a portfolio page explaining: “Initial model overfitted to rainy-day traffic; added spatial smoothing to reduce city-level variance by 22%.” The second candidate got the offer.

The portfolio must answer: What did you decide, and why?

Not “here’s my code,” but “here’s the tradeoff I faced, here’s how I resolved it.”

Include:

One A/B test teardown: design, threat assessment, decision impact
One modeling project with clear validation strategy (e.g., time-based split, synthetic control)
One example of data quality issue you identified and fixed

Avoid:

Kaggle-style projects without business context
Public datasets with no connection to mobility or behavior
Dashboards that show data but not decisions

A portfolio is not proof of technical skill — Didi assumes you can code. It’s proof you can think.

> 📖 Related: Didi PM hiring process complete guide 2026

How many projects should you list on your resume for Didi?

Three to four projects max — each must show end-to-end ownership. Didi’s interview loop includes a deep dive on one resume project, so every listed item must withstand 45 minutes of scrutiny.

In a 2025 debrief, a candidate listed six projects. The interviewer picked the weakest one — a basic cohort analysis from early career. The candidate couldn’t explain the statistical power of their test. The loop was downgraded from “strong hire” to “no hire.”

Quantity dilutes credibility. Depth builds it.

Each project should cover:

Business problem
Data constraints (e.g., missing driver GPS pings, rider surge bias)
Methodological choice (and alternatives rejected)
Validation approach
Measured outcome

For example:

“Designed counterfactual estimator for Didi Bike usage during subway outages, using synthetic control; estimated 23% uplift in idle city capacity, leading to dynamic pricing pilot”

This single bullet implies experimentation design, causal inference, domain awareness, and execution — all required for Didi DS work.

If you worked on a team, specify your role: “led analysis” vs. “contributed to model training.” Didi values ownership, not participation.

One more not X, but Y:

Not “worked on driver ETA model,” but “identified GPS drift in tunnel segments degrading ETA accuracy by 15%, implemented Kalman filter adjustment validated via OLS residuals, reducing MAE by 11%”

The second version shows diagnostic skill, technical action, and measurable correction.

Preparation Checklist

Quantify every impact: use percentages, time saved, revenue impact, or error reduction
Align projects with Didi’s domains: dispatch, pricing, safety, driver retention, rider growth
Use precise technical language: “propensity score matching” not “adjusted for bias”
Remove all fluff: “team player,” “strong communicator,” “passionate about data”
Include city-level or region-specific results if available — Didi operates under hyperlocal dynamics
Work through a structured preparation system (the PM Interview Playbook covers mobility data science case frameworks with real Didi debrief examples from 2023–2025)

Mistakes to Avoid

BAD: “Improved user engagement using machine learning”

This says nothing. What engagement metric? What model? What was the baseline? Hiring managers assume the worst: that you don’t know either.

GOOD: “Increased 7-day rider retention by 6.2% by re-ranking discovery feed using gradient-boosted CTR model, validated via two-week A/A test and two-city rollout”

This version specifies metric, method, validation, and deployment scope — all required for credibility.

BAD: “Used SQL and Python to analyze data”

This is table stakes. Every candidate does this. It’s like saying “used hands to type.” It adds zero signal.

GOOD: “Reduced SQL query latency from 18 minutes to 5.3 minutes by rewriting window functions and adding date partitioning, enabling daily automation of city performance reports”

This shows technical initiative with measurable impact — exactly what Didi looks for.

BAD: “Collaborated with product and engineering teams”

This is noise. Didi expects cross-functional work. What mattered was your decision — not that you attended meetings.

GOOD: “Convinced product team to delay feature launch after detecting interference in A/B test due to driver app update overlap, preventing invalid inference”

This shows judgment, risk assessment, and influence — leadership markers.

FAQ

Do Didi data scientists need to know mobility-specific metrics?

Yes. You must understand metrics like dispatch success rate, time-to-match, surge multiplier elasticity, and driver idle time. Not knowing these signals you’re a generalist — not a fit for Didi’s domain-heavy work. In a 2024 interview, a candidate confused CTR with ride acceptance rate and was immediately disqualified.

Should you include salary expectations in your resume?

No. Didi does not want salary in resumes. Discuss only during offer stage. Salaries for mid-level DS roles range from ¥480,000–¥720,000 annually, with senior roles exceeding ¥1M with stock. Mentioning it early signals desperation or misalignment.

How long should your resume be for a Didi data scientist role?

One page. No exceptions. Didi recruiters spend an average of 47 seconds per resume. Senior candidates with 8+ years may use two pages only if every line shows unique impact. In a 2025 HC, a two-page resume was rejected because the second page contained old internship details from 2016. Relevance decays fast.

Ready to build a real interview prep system?

Get the full PM Interview Prep System →

The book is also available on Amazon Kindle.