DEV Community

Saarathi Academy for Saarathi Academy

Posted on Edited on Originally published at linkedin.com

How to judge an AI course in Nepal before you pay

A good AI course in Nepal makes you build an AI app, test whether its answers are right, and put it online. Before you pay, read the week-by-week syllabus and look for those three things.

Many pages for AI courses in Nepal list tools: ChatGPT, LangChain, "agents". A list of tools tells you very little. The six questions below tell you much more, and you can ask them at any course in Nepal.

What should an AI course in Nepal teach first?

Python first, then SQL, then LLMs. LLMs are large language models, the kind of model behind ChatGPT, Claude and Gemini.

Look at what employers in Nepal ask for. An AI automation engineer post on the F1Soft careers site asks for strong Python, experience with LLM APIs from OpenAI, Anthropic and Google, and at least one RAG or agent framework. An AI/ML engineer post on merojob in 2025 asked for strong Python, machine learning algorithms and data preparation. Neither post asks for clever prompts.

If the first month has no Python, we would call it a prompting course. That can be useful for a manager or a marketer. It trains you for a different job than AI engineer, though, so check which one you are paying for.

Does the course test whether the AI's answers are right?

This question tells you the most about a course. Ask it plainly: "How do students check that their chatbot gives correct answers?"

A good answer sounds like this. Students write a set of test questions with the answers they expect. They run the app against that set and get a score. When they change a prompt, they run the set again and see if the score went up or down.

A weak answer is "we try it and see". Anyone can build a chatbot that looks fine in a demo. The skill companies need is knowing when it is wrong, and how often.

Can you take AI classes in Nepal on free tiers and your own laptop?

Paid API keys are billed in dollars, and many learners in Nepal do not have a dollar card. Ask which accounts you need and what you will pay for outside the course fee.

A course can run on free tiers today. The Gemini API has a free tier with free input and output tokens on some models. Google's page also says content sent on the free tier is used to improve its products. So never paste private company data or anyone's personal details into it.

Google Colab is free too, but its FAQ says GPU access on the free version is "heavily restricted". So also ask how much RAM your laptop needs, and which steps run in the cloud instead.

Does a project go live, and who checks it?

A notebook on your laptop is not a portfolio. Ask if your project ends up at a public web address that an employer can open. Also ask if it has a README that explains how it works and what each request costs.

Then ask who reviews your code, and how many students that person reviews. A mentor with a very large class cannot read every project line by line. Ask whether the mentor builds AI apps for work today, or only teaches.

Does it cover AI security?

An AI app that reads user input can be tricked. The OWASP Top 10 for LLM applications is a free list of the most common risks. Its first item in the 2025 version is prompt injection. That is when a user types text that makes your app ignore its own rules.

Ask if the course teaches you to defend against it. If the syllabus never mentions prompt injection, your first app may leak data or do things it should not.

If you are still deciding whether AI is the right field at all, our article on the future of AI and data science training in Nepal explains why these skills are in demand.

Does it promise a job or a salary?

Be careful with any course that promises a job or a salary number. A course cannot control what an employer decides. What a course can give you is skill, and honest feedback on real work.

How our AI course in Kathmandu answers these six questions

We teach AI engineering at Saarathi Academy in Old Baneshwor, Kathmandu, in person and online. Here is how our course answers each question, from its public syllabus.

  • What comes first: Python from zero in Weeks 1 and 2, SQL in Week 3, LLMs from Week 6.
  • Testing answers: students build a small test set of prompts and expected answers in Week 6, and measure RAG apps with RAGAS in Week 8. RAG means an AI that answers from your own documents.
  • Free tiers: class runs on free tiers of Gemini, Groq and OpenRouter. 16GB of RAM is recommended, and 8GB works with free Colab for the heavy steps.
  • Going live: in Week 12 you deploy a project online with FastAPI and Docker, with the cost of each request in its README.
  • Who checks: up to 10 students in the classroom per batch, and a few online. Mentors still do the job they teach.
  • Security: prompt injection and the OWASP LLM Top 10 come in Week 10.

It is 12 weeks, Monday to Friday, two hours a day. It is fast, so plan for two to three hours of practice every day on top of class. The course ends with a certification exam taken online at home. We do not promise a salary.

What can you try tonight?

Open the free Hugging Face LLM course and read chapter 1. It is free with no ads, and it shows how a transformer model works. It expects good Python, so if chapter 1 feels hard, that tells you to learn Python first.

Then hold any course you are thinking about against the six questions above. You can read our week-by-week AI Engineering syllabus and the page for our AI course in Nepal before you decide anything.

Common questions

Is an artificial intelligence course in Nepal the same as an AI engineering course?

Not always. The same name is used for short courses on using AI tools and for longer ones on building with them. Read the syllabus: an AI engineering course starts with Python and ends with an app you built, tested and put online.

What should an LLM course in Nepal cover besides prompts?

It should teach you to test whether the answers are right, with a fixed set of questions and a score. It should cover prompt injection from the OWASP Top 10 for LLM applications. It should also end with a project that runs at a public web address.

Can an AI course in Nepal guarantee a job?

No. Hiring is the employer's decision, so treat a job promise or a salary number in an advert as a warning sign. Ask instead what you will have built by the last week, and who reviewed it.

Top comments (0)