AIML Roadmap for BTech Students: 1st to 4th Year
Updated
01The 4-year plan at a glance
| Year | Learn | Build | Milestone by year-end |
|---|---|---|---|
| 1st year | Python, Git, calculus & linear algebra basics | Small Python scripts and CLI tools | Comfortable writing 100–200 line Python programs; GitHub account with a few repos |
| 2nd year | DSA, SQL, NumPy/pandas, probability & stats, classical ML | EDA reports, first scikit-learn models | Summer: first end-to-end ML project on GitHub with a README |
| 3rd year | Deep learning (PyTorch), NLP/CV, LLM APIs, RAG, deployment | A deployed deep learning or LLM app | Internship applications out; 2–3 solid portfolio projects |
| 4th year | One specialisation in depth (AI engineering, ML engineering or data science) | Final-year project that solves a real problem | Full-time applications and placement drives |
This runs alongside your university syllabus, not instead of it. See the CSE-AIML syllabus guide for what each semester teaches and the free resource we recommend for every subject.
021st year: Python, maths and good habits
Semester 1
- Python properly: variables, loops, functions, lists and dicts, files, errors. Harvard CS50P is free and excellent. Even if your college teaches C first, start Python on the side.
- Your college maths, understood visually: pair the calculus and matrices you are already studying with 3Blue1Brown's Essence of Calculus and Essence of Linear Algebra.
- Git and GitHub: make an account and push every program you write. A green contribution graph from year one is worth more than it sounds.
Semester 2
- Intermediate Python: classes and objects, modules, virtual environments, reading APIs with
requests. - Linux command line basics: navigating folders, running scripts, installing packages.
- Build: two or three small tools you actually use, such as an expense tracker, a CGPA calculator or a script that downloads data from a public API.
1st-year summer: start data structures in Python or C++ and practise a few problems a week. Rest matters too; don't burn out before 2nd year.
032nd year: DSA, data and your first ML model
Semester 3
- Data structures and algorithms: keep pace with your DSA subject and practise regularly. Most company hiring tests start here, AI roles included.
- NumPy and pandas: load a CSV, clean it, group it, plot it. The Data Handling section of the AI Engineer roadmap has the resources.
- SQL: learn it alongside DBMS. Joins, group by and window functions are used daily in data jobs.
Semester 4
- Probability and statistics: distributions, Bayes, hypothesis tests. StatQuest videos make this painless.
- Classical machine learning: regression, classification, trees and ensembles, clustering, model evaluation, all in scikit-learn. Follow the Machine Learning roadmap in order.
- Build: an exploratory data analysis report on a public dataset (an Indian one from data.gov.in is a nice touch), then a Titanic or house-price model.
2nd-year summer milestone: your first end-to-end ML project on GitHub. Pick something like customer churn prediction or a crop disease classifier: clean data, train and compare models, explain the results, and write a README with screenshots. If you can, put a small Streamlit or Gradio demo online. Ideas with datasets are in AIML project ideas.
043rd year: deep learning, LLMs and the internship hunt
Semester 5
- Deep learning with PyTorch: neural networks, backpropagation, CNNs, training loops. Karpathy's Zero to Hero and fast.ai are the best free routes; the Deep Learning and PyTorch sections of the AI Engineer roadmap link both.
- Pick a direction: NLP and transformers, or computer vision. You will have a college elective in one of them anyway.
- Software skills: FastAPI, Docker and basic cloud deployment, so your models can run somewhere other than a notebook.
Semester 6
- LLMs and AI engineering: LLM APIs, prompt and context engineering, embeddings, RAG and simple agents. This is the most in-demand skill set for new AI roles; the AI Engineer roadmap covers it step by step.
- Build: a deployed LLM app, such as a PDF Q&A chatbot with citations over your college notes, plus one deep learning project (e.g. a fine-tuned BERT or a transfer-learning image classifier).
- Internship applications: most 3rd-year students look for a summer internship in this semester. Keep a one-page resume with your best 2–3 projects and links. Browse openings on our jobs board and the companies that hire for AI/ML.
3rd-year milestone: an internship (or, if that doesn't work out, a serious open-source contribution or a research project with a professor) and a GitHub profile with 2–3 projects you can explain end to end.
054th year: specialise, ship a final project, apply
Pick one direction for depth and follow its roadmap:
| If you enjoy… | Aim for | Follow |
|---|---|---|
| Building products with LLMs, RAG and agents | AI Engineer | AI Engineer roadmap |
| Training, deploying and monitoring models | ML Engineer | Machine Learning roadmap |
| Statistics, experiments and business questions | Data Scientist / Analyst | Data Scientist roadmap |
- Semester 7: start your final-year project early and make it real: real users, real data, deployed, with an evaluation. Add a second skill on top, like MLOps (experiment tracking, CI) or LLM evaluation.
- Semester 8: finish and document the project, write a short blog post or demo video about it, and apply widely: campus placements, off-campus roles, and startups.
Not sure which role suits you? Compare them in AI engineer vs ML engineer and AI engineer vs data scientist, and see what each pays in AI/ML engineer salary in India.
06If you are starting late
Already in 3rd year with little done? Don't try to cover everything. Do Python and pandas fast (3–4 weeks), classical ML (4–6 weeks), then go straight to one LLM project and one ML project and deploy both. Two working, well-explained projects beat ten half-finished courses. The main AIML roadmap shows the shortest path.
07Frequently asked questions
What should I learn in 1st year of BTech AIML?
Python, Git and the maths you are already studying (calculus and linear algebra), understood well. Build a few small Python tools and push them to GitHub. You don't need machine learning yet.
When should I start machine learning in BTech?
Around the 4th semester or the 2nd-year summer, once you are comfortable with Python, pandas and basic probability. That puts you a semester ahead of the ML subject in most syllabi.
Is DSA needed for AIML students?
Yes. Most companies, including for AI and ML roles, run coding tests based on data structures and algorithms, and DSA is part of every CSE-AIML syllabus.
How many projects do I need before applying for an internship?
Two or three good ones are enough: complete, on GitHub with a clear README, and ideally with a live demo. Quality and how well you can explain them matter more than the count.
Should I learn generative AI or classical ML first?
Learn the basics of classical ML first (it takes a few weeks), then move to LLMs and generative AI. The fundamentals make RAG, fine-tuning and evaluation much easier to understand.