About half
Share of 1,355 US data analyst job postings that mention SQL, with Python in about a third and machine learning in 14%. Querying and analysis come before model training
365 Data Science analysis of Glassdoor US postings, April 2026
$35-$55/hr
Typical US hourly range for paid data science or ML intern roles at large tech and finance employers (varies by city and year of study)
Levels.fyi and employer intern cohorts, synthesised June 2026
4 proofs
What hiring managers weight most: SQL + one dataset story, one notebook with clear metrics, basic stats/ML literacy, and communication in a short write-up
Studojo data hiring-manager interview synthesis, 2025 to 2026
Job boards bundle data analyst intern, data science intern, ML intern, AI research intern, and business intelligence intern into one hype bucket. In 2026, most undergrad and master's interns spend time on definable data work: writing SQL, fixing datasets, building dashboards, running A/B analysis, labeling and evaluating model outputs, or wiring APIs into product features. Training foundation models from scratch is rare at intern level outside a handful of research labs.
GenAI shifted the stack, not the entry bar. Employers want interns who can measure quality, debug bad outputs, and ship small features with guardrails. That is closer to analytics plus software hygiene than to Hollywood AI.
Key insight: Optimize for data credibility first. AI fluency is a layer on top of SQL, Python, and clear metrics.
What "data/AI" intern roles actually do (illustrative mix of work, %)
Research intern ≠ corporate intern. University labs and FAANG research programmes expect coursework and papers. Product companies expect shipping and communication. Data ops is a valid foot in the door. Cleaning data and building pipelines teaches how models fail in production. Do not dismiss it if paid and mentored. Title decoder: "Data analyst intern" → SQL + dashboards. "Data science intern" → notebooks + experiments. "ML intern" → features + evaluation. "AI intern" → read the bullets; often product analytics with LLM APIs.
Recruiters skim for SQL depth (joins, window functions, sane data models), Python in notebooks (pandas, matplotlib or plotly), basic statistics (hypothesis tests, confidence, train/test intuition), and one portfolio project with a metric that moved. GitHub with readable README beats twenty Kaggle medals with no narrative.
For GenAI-facing roles, show prompt evaluation, RAG basics, or failure analysis on a small corpus. You do not need to fine-tune a 70B model. You need to show you can define success, measure it, and iterate.
Skill priority in data/AI intern job posts (index 0 to 10)
Key insight: One end-to-end project (question → data → analysis → decision) beats listing sklearn on your resume without context.
"I hire data interns who explain one decision their analysis changed. Most portfolios only show plots."
Analytics manager, B2B SaaS (Studojo community, 2025)LeetCode-heavy ML interviews are niche. Many data intern screens are SQL + case study + take-home, not hard competitive programming. Cloud certs are optional. BigQuery or Snowflake exposure helps. A project that queries a warehouse beats a badge alone. 90-day skill stack: Week 1-4: SQL on a real schema. Week 5-8: one Kaggle or public dataset with a written report. Week 9-12: small app or dashboard plus 5-slide presentation. Optional: one LLM eval notebook with labeled examples.
In the United States, large tech and finance data interns often land roughly $35 to $55 per hour in major metros, with housing stipends sometimes added. Smaller startups vary from competitive hourly to low stipends. Unpaid data internships at for-profit firms remain a red flag (see Studojo's unpaid internship report).
In India, structured tech and GCC data interns often see monthly stipends roughly in the ₹25K-₹80K band for summer programmes, highly employer-dependent. Startups may offer less cash plus project ownership. Always confirm in the offer letter.
Illustrative monthly pay index (US tech/finance vs India product/GCC, index 0 to 25)
Key insight: Paid data interns exist in volume at employers with real data teams. If everyone on the team is paid except interns, question the setup.
Remote US intern from India. Confirm currency, hours, and tax. Some US employers hire contractors with different rules than US campus interns. PhD vs undergrad bands differ. Research labs and quant firms pay premiums. Do not compare your offer to a friend's without matching level and city. Negotiate scope, not only stipend: Ask about mentor, warehouse access, presentation to leadership, and return offer history.
Large tech, fintech, and e-commerce run the biggest cohorts: product analytics, risk, search, ads, and platform data. Banks and asset managers hire quant-leaning and analytics interns. GCCs in India hire data engineering and BI interns for global stacks. Health, retail, and logistics hire when they have centralized data teams, not when "AI" is a press release only.
Consulting and agencies hire analytics interns for client dashboards. AI startups hire if you can ship evaluations and prototypes, not because you watched a transformer lecture.
Key insight: Target employers with a named data org chart. "AI-first" marketing without data job postings is a warning.
Capstone and university labs count. Professor research with a publication or shipped artifact helps research intern screens. Hackathons are side doors. Winning a data track with a reproducible repo sometimes skips the queue for startups. Search strings: "data analyst intern," "analytics intern," "data science intern summer 2026," "ML intern," plus company careers filter. Add city if you target India GCC or US hub.
Typical loops: SQL screen (live or timed), statistics and product sense questions, take-home analysis with a deck, and behavioral questions on ambiguity. ML-leaning roles may add Python coding or model metric questions. Rarely will undergrads face deep architecture proofs unless applying to elite research programmes.
Take-homes should be time-boxed (2 to 4 hours of real work). If an employer assigns a week of unpaid labor, decline or negotiate. Present insights, not notebook dumps.
Reach data hiring managers directly
Studojo Outreach helps you message analytics and ML leads with one project link before you are buried in a generic intern queue.
Try Studojo Outreach →Key insight: Interviewers reward clarity: metric, method, limitation, next step. Practice saying "I don't know, but I would test X."
Explain tradeoffs aloud. Why median vs mean, why logistic vs linear, why you dropped outliers. Junior hires win on judgment narration. GenAI take-homes often test evaluation. Design rubrics, human rating samples, and failure buckets. That is real 2026 work. Take-home template: Problem → Data quirks → Analysis → Chart → Recommendation → What I'd do with two more weeks.
Days 1-20: finish SQL + one portfolio project with a README and slides. Days 21-40: apply to 15 tailored roles (5 large tech/finance, 5 India GCC or product, 5 startups). Days 41-60: ten outreaches to data managers with your project link; mock SQL twice a week.
Track screens per ten tailored applies. If only startups reply, tighten dashboards. If only GCC replies, emphasize SQL and pipeline hygiene. Do not spray "AI enthusiast" resumes.
Summary insight: Entry-level data and AI hiring is a data credibility game with an AI accent. Build the base, then add LLM literacy with measured projects.
"The intern who got the return offer explained one dashboard that changed a team's sprint priority. The others had pretty plots."
Head of data, consumer marketplace (Studojo community, 2025)Avoid fake AI projects. Wrappers around ChatGPT with no evaluation metric hurt trust. Show evals, costs, and failure modes. Return offers follow communication. Interns who present weekly to mentors convert more than interns who only code in silence. Portfolio must-haves: One SQL repo, one notebook with business recommendation, one slide deck under 6 pages, LinkedIn headline that names your stack honestly.
Most AI interns do data work. SQL, cleaning, dashboards, and evaluation beat frontier training at entry level. Read the job bullets, not only the title. Ship one defensible project. Question, data, method, metric, recommendation. That narrative beats tool lists and buzzwords. Target employers with real data teams. Large tech, finance, GCCs, and verticals with centralized analytics hire paid cohorts. Skip vague "AI-first" posts with no data org. Practice SQL and communication. Screens are often SQL + case + take-home. Present tradeoffs clearly; time-box unpaid homework and walk if abused. What do data and AI interns actually do in 2026?
Most data and AI interns do data work, not frontier model training. In the report's illustrative mix, analytics and BI (SQL, dashboards) is 34% of the work, data engineering lite is 22%, applied ML is 18%, and GenAI product support is 14%. Research-style ML is rarer at intern level, at 6%.
What skills do I need for a data science or AI internship?
SQL and Python in notebooks come first, followed by statistics and experimentation, visualization, basic ML, and LLM or API familiarity. In the report's skill index, SQL scores 9.5 out of 10 and Python 9.2. One end-to-end project (question, data, analysis, decision) beats listing sklearn on your resume without context.
How much do data and AI interns get paid?
In the US, large tech and finance data interns often land roughly $35 to $55 per hour in major metros. In India, structured tech and GCC data interns often see roughly ₹25K to ₹80K per month for summer programmes, highly employer-dependent. Confirm the number in your offer letter, and treat unpaid or stipend-TBD roles at for-profit firms as a red flag.
Do I need to build or fine-tune an LLM to get an AI internship?
No. Training foundation models from scratch is rare at intern level outside a handful of research labs. For GenAI-facing roles, show prompt evaluation, RAG basics, or failure analysis on a small corpus, and prove you can define success, measure it, and iterate.
What does a data or ML intern interview look like?
A typical loop is a SQL screen, statistics and product sense questions, a take-home analysis with a deck, and behavioral questions on ambiguity. ML-leaning roles may add Python coding or model metric questions. Take-homes should be time-boxed to 2 to 4 hours of real work, so decline or negotiate if an employer assigns a week of unpaid labor.
How can I break into data and AI internships in 60 days?
Spend days 1 to 20 finishing SQL and one portfolio project with a README and slides. Use days 21 to 40 to apply to 15 tailored roles across large tech or finance, India GCC or product firms, and startups. In days 41 to 60, send ten outreaches to data managers with your project link and run mock SQL twice a week.
Land a paid data intern role with proof.
Studojo helps you find structured data and analytics internships and reach hiring managers with a project link, not a generic AI resume.