Requiring deep-learning experience for a role that is 80% SQL and cleaning is the first thing to go wrong in most data science intern job descriptions — and the second is that they list requirements without ever describing the work. A student reading one cannot picture a Tuesday, so they either do not apply or apply to forty postings indiscriminately, and you get volume without fit.
The template below is the opposite shape. It opens with what this data science intern would actually do — clean and explore a real internal dataset, among other things — states what success looks like by the end, prices the role as a number, and sets out a selection process built around a three-hour exercise on a real, messy dataset: state a question, answer it, and write down what would falsify your answer. Everything in square brackets is yours to fill in; everything else is ready to post.
The template — copy this
Data Science Intern — [City] =============================== ABOUT THE INTERNSHIP [Company name] is hiring a Data Science intern for a 3-month internship based in [City]. This is a working internship on live projects — not a training programme and not shadowing. In your first week you will get access to a real dataset, meet the team who owns it, and produce one honest first look. You will have a named mentor, a scoped project, and a review at the midpoint and at the end. WHAT YOU WILL DO • Clean and explore a real internal dataset • Build and evaluate a baseline model against a stated metric • Document assumptions, leakage risks and limitations • Present findings to a non-technical stakeholder WHAT SUCCESS LOOKS LIKE One analysis or model with an honest evaluation the business acted on or consciously shelved. We will tell you at the start exactly what we will be assessing, and you will get written feedback at the midpoint so nothing at the end is a surprise. WHAT WE ARE LOOKING FOR • B.Tech, B.Sc/M.Sc (Statistics, Maths, CS, Economics), MCA or M.Tech. Final-year students and recent graduates are both welcome. • Working familiarity with python (pandas, NumPy) • Working familiarity with statistics and hypothesis testing • Working familiarity with sQL • Working familiarity with scikit-learn fundamentals • Working familiarity with communicating findings clearly • The ability to ask a clear question when you are stuck, rather than staying stuck WHAT WE OFFER • A stipend of [stipend] per month, paid monthly • A 3-month internship starting [start date] • A named mentor and a weekly one-to-one, not an unassigned desk • A certificate and a written reference reflecting your actual work • Access to real internal data, under our data-handling policy, rather than a sample set • A pre-placement offer for interns who meet the bar — this is a real path, not a maybe SELECTION PROCESS • Application review — we read every application and reply either way • A 20-minute screening call about your background and what you want from the internship • A practical task: a three-hour exercise on a real, messy dataset: state a question, answer it, and write down what would falsify your answer. It is timeboxed, we will tell you what we are looking for, and someone will read it and give you feedback either way. • A final conversation with the team you would be joining • Decision and offer within a week of the final round HOW TO APPLY Apply through this listing by [application deadline]. Attach your CV and, if you have one, a link to an analysis or model you have built and can talk through. Coursework counts if it was yours. [Company name] is an equal-opportunity employer; we assess every application against the same criteria, and we reply to everyone who reaches a screening call.
What this data science intern will actually do
These are the responsibilities the template uses. They are deliberately concrete: an intern can picture the work, and you can assess against them at the end.
- Clean and explore a real internal dataset
- Build and evaluate a baseline model against a stated metric
- Document assumptions, leakage risks and limitations
- Present findings to a non-technical stakeholder
What to screen for in a data science intern
Do they state what would falsify their conclusion? That one question separates candidates for this role far more reliably than the CV does, because almost nobody applying has directly relevant work history. The practical way to ask it is a three-hour exercise on a real, messy dataset: state a question, answer it, and write down what would falsify your answer — timeboxed, reviewed, and with a line of feedback back either way.
- Python (pandas, NumPy)
- The one to insist on. Depth here beats a passing acquaintance with everything else on this list.
- Statistics and hypothesis testing
- Working familiarity is enough at intern level — this is learnable on the job if the first item is genuinely there.
- SQL
- Working familiarity is enough at intern level — this is learnable on the job if the first item is genuinely there.
- scikit-learn fundamentals
- Working familiarity is enough at intern level — this is learnable on the job if the first item is genuinely there.
- Communicating findings clearly
- Working familiarity is enough at intern level — this is learnable on the job if the first item is genuinely there.
Why this JD converts
- It commits to a real internal dataset. Data-science interns are routinely given a Kaggle set and a laptop; the ones worth hiring can tell the difference before they accept.
- It asks for documented assumptions and leakage risk, which is the actual failure mode of junior data work — and it tells the candidate you know that.
- It includes presenting to a non-technical stakeholder in the responsibilities, so communication is assessed as part of the job rather than bolted on as a "good communication skills" line nobody screens for.
What employers get wrong hiring data science interns
- Requiring deep-learning experience for a role that is 80% SQL and cleaning
- You screen out the candidates who would have been good at the actual work and hire someone who is bored by it within a month.
- Not saying whether the data is real
- Ambiguity here reads as "sandbox project". Say plainly whether the intern touches production data and under what controls.
- Conflating data science with data analytics in the title
- Two different candidate pools with different expectations. Mis-titling gets you applications from both and the right fit from neither.
What to pay a data science intern
For a three-month, in-office internship at a mid-sized company, the benchmark band for this role in a tier-1 metro is ₹18,000–₹54,500 a month, and roles priced in the upper-middle of that band fill fastest. Adjust down for tier-2 cities and remote work, up for large enterprises and GCCs. State the number in the JD — "as per company norms" reads as "low", and the strongest applicants filter it out before you ever see them. For this role specifically, the thing candidates weigh against the stipend is one analysis or model with an honest evaluation the business acted on or consciously shelved — a credible answer there is worth several thousand rupees a month in how fast the role fills.
Who to open this data science intern role to
The template uses a deliberately wide eligibility line: B.Tech, B.Sc/M.Sc (Statistics, Maths, CS, Economics), MCA or M.Tech. Widen it rather than narrowing it. Long undifferentiated requirement lists measurably suppress applications from candidates who would have succeeded, and the effect is strongest among exactly the candidates most internship programmes say they want to reach. B.Tech and M.Tech, plus M.Sc and B.Sc in Statistics, Mathematics and Economics — the non-engineering statistics pool is consistently under-recruited in India and is often better trained in inference than the CS pool.
Frequently asked
Should a data science intern JD require machine learning experience?+
Only if the internship is actually about modelling. Most data-science internships are eighty percent cleaning, joining and questioning data, and requiring deep-learning experience for that work screens out the candidates who would have been good at it while attracting ones who will be bored within a month. Describe the actual split in the JD.
How do we say whether the intern will touch real data?+
Plainly, and with the control. "You will work on a real internal dataset, under our data-handling policy, with access limited to the tables your project needs" answers the question a good candidate is silently asking. Ambiguity is read as "this is a Kaggle project", and the candidates who can tell the difference are the ones you want.
Should the title say data science or data analytics?+
Whichever the work actually is — they draw different candidate pools with different expectations. Mis-titling a heavily SQL-and-dashboard role as data science gets you applications from both pools and a good fit from neither, and the mismatch surfaces in week two rather than at offer stage.
Can we use this data science intern template commercially?+
Yes — copy it, edit it and post it anywhere, with no attribution required. On JD length and structure generally, the master internship job description template covers what applies to every role; this page covers what is specific to hiring a data science intern.