Data Science & Analytics Internship

Turn a messy dataset into a decision somebody can act on.

Analysis that ends in a recommendation rather than a chart. You will clean real, awkward data, find something genuinely non-obvious in it, and present it so a decision-maker can act — the skill that separates an analyst from a dashboard operator.

Beginner-friendly4 weeks · remote4 briefs · complete any 2Free to join

What you'll be able to do

Clean and reconcile untidy real-world data
Do exploratory analysis that produces a finding, not just plots
Build a dashboard someone would actually open twice
Write up an analysis with a recommendation attached

Who it suits

Students who can write basic Python and want to work with real data rather than a cleaned teaching set.

What you'll work with

Python
pandas
NumPy
Matplotlib/Seaborn
SQL
Streamlit

The project briefs

Complete any 2 of these 4to finish the programme. Assessment criteria are published here so you know what “good” means before you start.

Real-World Data Cleaning Pipeline

~8h estimated

Turn a genuinely messy public dataset into something analysable, reproducibly.

Pick a dataset with real problems — missing values, inconsistent categories, duplicates, mixed date formats. Write the cleaning as a reproducible script, not as manual edits.

Assessed against

Every cleaning decision documented with its justification
Rerunnable from raw data with one command
Before/after data quality summary
Rows dropped are counted and explained

Suggested datasets

data.gov.in
Kaggle raw datasets
municipal open-data portals

You'll finish with

GitHub repo with the pipeline
Data quality report

Exploratory Analysis with a Finding

~10h estimated

Find something in the data that a stakeholder did not already know.

Explore systematically rather than plotting at random. The deliverable is a claim with evidence — and a check that the pattern is not an artefact of how the data was collected.

Assessed against

A stated hypothesis and the evidence for it
At least one initially-promising pattern you disproved
Charts labelled well enough to stand alone
A plain-language summary of the finding

You'll finish with

Notebook
Two-page findings summary

Interactive Analytics Dashboard

~12h estimated

Build a dashboard that answers a specific recurring question.

Start from the question, not the chart library. Every element must earn its place — if it does not change a decision, cut it.

Assessed against

The question it answers, stated on the dashboard
Meaningful filters, not decorative ones
Loads in under three seconds
Deployed and publicly reachable

You'll finish with

Live deployed URL
GitHub repo

Time-Series Forecast with Honest Error Bars

~12h estimated

Forecast a real series and be truthful about the uncertainty.

Establish a naive baseline first — seasonal-naive beats a lot of sophisticated models. Backtest properly with a rolling origin; a random split leaks the future into the past.

Assessed against

A naive baseline the model must beat
Rolling-origin backtesting, not a random split
Prediction intervals, not just point estimates
A note on when this forecast should not be trusted

Suggested datasets

Retail sales, energy demand, air quality, or ridership series

You'll finish with

GitHub repo
Backtest results with intervals

Free to apply and free to complete. The completion certificate is optional.

Other domains

AI Agents & Automation

Build autonomous agents that use tools, not chatbots that answer questions.

Generative AI & LLMs

Go past prompt tinkering into retrieval, evaluation and cost control.

Machine Learning

Classical ML done properly — including the evaluation everyone skips.

Ready to Transform Your Enterprise with AI?

Book a free 30-minute consultation with our AI architects. Discover how agentic AI can automate your workflows and multiply your team's output.

No commitment required · Free architecture review · NDA protected

Data Science & Analytics Internship

Turn a messy dataset into a decision somebody can act on.

Beginner-friendly4 weeks · remote4 briefs · complete any 2Free to join

What you'll be able to do

Clean and reconcile untidy real-world data

Do exploratory analysis that produces a finding, not just plots

Build a dashboard someone would actually open twice

Write up an analysis with a recommendation attached

Who it suits

Students who can write basic Python and want to work with real data rather than a cleaned teaching set.

The project briefs

Complete any 2 of these 4to finish the programme. Assessment criteria are published here so you know what “good” means before you start.

Real-World Data Cleaning Pipeline

~8h estimated

Turn a genuinely messy public dataset into something analysable, reproducibly.

Pick a dataset with real problems — missing values, inconsistent categories, duplicates, mixed date formats. Write the cleaning as a reproducible script, not as manual edits.

Assessed against

Every cleaning decision documented with its justification
Rerunnable from raw data with one command
Before/after data quality summary
Rows dropped are counted and explained

Suggested datasets

data.gov.in
Kaggle raw datasets
municipal open-data portals

You'll finish with

GitHub repo with the pipeline
Data quality report

Exploratory Analysis with a Finding

~10h estimated

Find something in the data that a stakeholder did not already know.

Explore systematically rather than plotting at random. The deliverable is a claim with evidence — and a check that the pattern is not an artefact of how the data was collected.

Assessed against

A stated hypothesis and the evidence for it
At least one initially-promising pattern you disproved
Charts labelled well enough to stand alone
A plain-language summary of the finding

You'll finish with

Notebook
Two-page findings summary

Interactive Analytics Dashboard

~12h estimated

Build a dashboard that answers a specific recurring question.

Start from the question, not the chart library. Every element must earn its place — if it does not change a decision, cut it.

Assessed against

The question it answers, stated on the dashboard
Meaningful filters, not decorative ones
Loads in under three seconds
Deployed and publicly reachable

You'll finish with

Live deployed URL
GitHub repo

Time-Series Forecast with Honest Error Bars

~12h estimated

Forecast a real series and be truthful about the uncertainty.

Establish a naive baseline first — seasonal-naive beats a lot of sophisticated models. Backtest properly with a rolling origin; a random split leaks the future into the past.

Assessed against

A naive baseline the model must beat
Rolling-origin backtesting, not a random split
Prediction intervals, not just point estimates
A note on when this forecast should not be trusted

Suggested datasets

Retail sales, energy demand, air quality, or ridership series

You'll finish with

GitHub repo
Backtest results with intervals

Free to apply and free to complete. The completion certificate is optional.