In the first four weeks of a data science course in Nepal, learn to handle data, not to build models. That means Python basics, then Pandas next to Excel, then SQL, in that order. Machine learning can wait.
This is the order we teach at Saarathi Academy in Old Baneshwor, Kathmandu. Below is each week, what goes in it, and why it comes where it does. You can follow the same order on your own.
Why not start with machine learning?
It is tempting to open a machine learning video first. It feels like the real thing. It is the wrong place to start.
Entry level analyst work in Nepal is mostly cleaning data and preparing reports. A Nepali job portal's guide for data analysts says this plainly, and it tells candidates to expect interview questions on Excel, SQL, Power BI and data cleaning. A data analyst listing on merojob from a Kathmandu finance company asked for Excel pivot tables, Power BI dashboards, SQL and Python.
There is a second reason. A model trained on messy data gives wrong answers with full confidence. If you cannot clean a table, you cannot tell when your model is lying to you. So in our course, machine learning starts in Week 9, after the data skills are solid.
Week 1: Python basics, using what you already know
Week 1 is plain Python. Variables, numbers and text, if decisions, loops, functions, lists and dictionaries. You also set up Python, Jupyter, VS Code and Git, and make your first GitHub commit.
The useful trick is to rewrite spreadsheet formulas you already know. Write SUM, IF, VLOOKUP and COUNTIF in Python. You already know what the right answer is, so you can check your own code.
ChatGPT and Copilot can draft code from Week 1. The rule is simple: you check every line they write. If you cannot explain a line, you do not use it.
One warning if you study alone. The official Python tutorial says it is written for people who already program, not for complete beginners. If you are new, the free freeCodeCamp Python lessons are a gentler start.
Week 2: files, errors and NumPy
Week 2 is about real files and real errors. You read and write CSV files with plain Python. You learn to read a traceback, the error message Python prints when something breaks.
Then you build a small CSV analyser with no library at all. It is slower to write, and that is the point. When you later use Pandas, you know what it is doing for you.
The week ends with NumPy, a library for fast maths on whole columns of numbers. You see why arrays beat lists when the data gets large.
Week 3: Pandas and Excel side by side
Pandas is the Python library analysts use most for tables. In Week 3 you load CSV and Excel files, filter rows, group and total them, merge two tables, and fix missing values and duplicates.
We teach Excel in the same week, on purpose. You build a pivot table, an XLOOKUP and a small Excel dashboard. Then you build the same thing in Pandas. Many offices in Nepal still send reports in Excel, so you will need both.
Cleaning is where Nepal data gets its own problems. Office files often store dates in Bikram Sambat. Some BS months have 32 days, so a normal date parser rejects those rows. You learn to handle these dates before they break a report.
Week 4: SQL and databases
SQL is the language for asking questions of a database. In Week 4 you set up PostgreSQL and write SELECT, WHERE, GROUP BY, HAVING and JOIN. Then you learn CTEs, which are named steps inside a query, and window functions like ROW_NUMBER, RANK and LAG.
Why does SQL come after Pandas and not before? Because by Week 4 you have already grouped and merged tables in Pandas. GROUP BY in SQL is the same idea as groupby in Pandas, written a different way.
You also run SQL from Python and learn to avoid SQL injection, a common attack on badly written queries. The week ends with one question answered from four places: a CSV, an Excel file, a JSON API and Postgres. At work, data rarely sits in one place.
Will my old laptop be enough?
For our course you need your own laptop with at least 8GB of RAM. Every tool has a free version, so you do not pay for any account. Power BI Desktop, which comes in Week 8, runs only on Windows. On a Mac you need your own Windows setup before then.
If your laptop is older, you can still start tonight. Google Colab runs Jupyter notebooks in your browser, free, with nothing to install. Its free resources are not guaranteed, and an idle notebook times out, so save your work often.
What should I check before choosing a course?
Look for this same order in any syllabus you read. Python before libraries, data cleaning before charts, SQL before machine learning. If a course opens with deep learning in Week 1, ask what you will do with messy data.
We wrote a longer guide on how to choose data science training in Kathmandu, with the questions to ask before you pay.
What to try tonight
Open a new notebook in Colab. Then work through the first two pandas getting started tutorials: what kind of data pandas handles, and how to read and write tabular data. Use a CSV of your own monthly spending instead of their sample file. When you can answer "what did I spend most on?" in one line of Pandas, you are ready for Week 3.
Our full 12 weeks are on the Data Science and Machine Learning course page, and every week is in the public syllabus.
Top comments (0)