AI Jobs in India for Freshers: Roles, Skills and 3 Portfolio Projects

By Ravindra Bagale, trainer · LinkedIn · Last updated: 19 October 2026

Generative AI course · Chapter 51 · Estimated reading time: 13 minutes

1. Suppose Neha finishes this course

Suppose Neha is a B.Com graduate in Nashik. She finished this course, she can write good prompts, and she built a PDF chatbot. She opens Naukri and LinkedIn and types "AI jobs for freshers". She sees hundreds of posts with titles like "AI Engineer", "Prompt Engineer", "GenAI Developer", "AI Data Analyst" and "AI Trainer".

Neha has three questions:

1. Which of these titles fit what I can do today?
2. What do I learn next, in what order?
3. How do I prove my skills when I have no work experience?

One rule answers all three: employers hire freshers for what they can show, not for what they have watched. So the plan is: pick one role, learn only what that role asks, and build proof for every skill you list.

2. Five AI entry roles in plain words

Two things to notice:

• The same title means different jobs. One company's "Prompt Engineer" writes marketing prompts; another expects Python. Always read the skills list, not the title.
• Every role needs a domain. Accounts, HR, sales, healthcare or teaching. Your B.Com, BSc or work experience is an advantage, not a gap.

3. Worked example 1: 3 role roadmaps

A roadmap is the order in which you learn. Keep it short. Each step ends with something you can show.

Roadmap A: AI-enabled data analyst (8 weeks, no heavy coding)

1. Weeks 1 and 2: Excel formulas with AI help (the Excel formulas chapter), and check each formula by hand.
2. Weeks 3 and 4: SQL basics on a free sample database.
3. Weeks 5 and 6: one BI tool (Power BI or Looker Studio) and one dashboard.
4. Weeks 7 and 8: a sales report project with an AI-written summary that you fact-check.

Roadmap B: prompt engineer / AI content specialist (6 weeks)

1. Weeks 1 and 2: the prompting chapters (role, task, context, format; few-shot; structured output).
2. Weeks 3 and 4: build a library of 25 tested prompts for one domain, each with an expected output.
3. Week 5: hallucination checks and privacy rules (what never to paste).
4. Week 6: a before-and-after case study: weak prompt, better prompt, measured result.

Roadmap C: GenAI developer (12 weeks, Python)

1. Weeks 1 to 3: Python basics and your first API call (the Python and API chapters).
2. Weeks 4 to 6: embeddings, RAG and a vector database.
3. Weeks 7 to 9: an agent with tools and guardrails, in Python.
4. Weeks 10 to 12: deploy one app (AWS Bedrock or a small server) and write a clear README.

A prompt to make your own roadmap:

I am a fresher in India with a [your degree] degree. I want a [role] job.
I can already do: [3 things you can show].
Make an 8-week roadmap, 1 hour a day on weekdays and 3 hours on Sunday.
End every 2 weeks with one thing I can show an employer.
Use only free resources and say which ones are official.

Expected output (yours will be similar, not the same): a week-by-week table with topics, a free resource for each, and a "proof" column such as "dashboard link" or "GitHub repo". Check that every resource link opens and is official.

4. Worked example 2: a portfolio of 3 projects

You do not need 10 projects. You need 3 that match your role, each with a short README, a screenshot or a 60-second demo video, and the numbers you checked. Build them from this course:

A README template that recruiters can read in one minute:

# HR Policy Chatbot (RAG)
Problem: employees ask the same leave questions every day.
What it does: answers only from the company PDF and shows the page it used.
How I built it: Python, a free embedding model, Chroma, one chat model API.
How I tested it: 10 questions with known answers; 9 correct, 1 fixed by better chunking.
What I would do next: add login, and log every question.
Demo: [60-second video link]

Expected output: when you ask AI to review this README ("Act as a hiring manager. What is unclear?"), it usually asks for real numbers and a demo link. Add them. Use dummy company data only, never a real employer's files.

5. Worked example 3: compare a job post with your proof (Python)

This small script reads a job post, finds the skills it asks for, and checks which ones you can prove. The job post and Neha's skills are made up for learning.

# Compare the skills a job post asks for with the skills you can prove.
# The job post and Neha's skills are made up for learning.
job_post = """Junior GenAI Developer (fresher). Must have: Python, prompt engineering,
REST APIs, Git. Good to have: RAG, vector database, AWS, n8n. Show 2 projects on GitHub."""

ASK = ["python", "prompt engineering", "rest apis", "git", "rag", "vector database", "aws", "n8n"]
MUST = {"python", "prompt engineering", "rest apis", "git"}

mine = {  # skill -> proof you can show (empty text = no proof yet)
    "python": "bill calculator repo on GitHub",
    "prompt engineering": "25-prompt library in a PDF",
    "rag": "HR policy PDF bot, demo video",
    "n8n": "email-to-sheet workflow, screenshots",
    "git": "",
}

text = job_post.lower()
asked = [s for s in ASK if s in text]
proved = [s for s in asked if mine.get(s)]
print(f"Skills asked: {len(asked)}  proved: {len(proved)}  match: {round(100 * len(proved) / len(asked))}%")
for s in asked:
    tag = "MUST" if s in MUST else "good"
    status = "proof: " + mine[s] if mine.get(s) else ("no proof yet" if s in mine else "missing")
    print(f"  [{tag}] {s:<19} {status}")
gaps = [s for s in asked if s in MUST and not mine.get(s)]
print("Fix first (must-have, no proof):", ", ".join(gaps) if gaps else "none")

Real output (19 October 2026):

Skills asked: 8  proved: 4  match: 50%
  [MUST] python              proof: bill calculator repo on GitHub
  [MUST] prompt engineering  proof: 25-prompt library in a PDF
  [MUST] rest apis           missing
  [MUST] git                 no proof yet
  [good] rag                 proof: HR policy PDF bot, demo video
  [good] vector database     missing
  [good] aws                 missing
  [good] n8n                 proof: email-to-sheet workflow, screenshots
Fix first (must-have, no proof): rest apis, git

Neha's next two weeks are now clear: put her projects on GitHub with proper commits (that proves Git), and add one small REST API call to her chatbot. She does not need AWS for this job yet. Run the script on 5 real job posts for your role; the skills that come up again and again are your learning list.

6. A LinkedIn headline that says the right thing

A recruiter searches LinkedIn by role and skill words. A headline like "Fresher | Looking for opportunities" matches nothing.

Formula: target role + 2 or 3 proven skills + one proof.

Weak:   Fresher | B.Com | Looking for job
Better: AI-enabled Data Analyst | Excel, SQL, Power BI | Built a sales dashboard with AI summaries
Better: GenAI Developer (fresher) | Python, RAG, n8n | Built a cited-answer HR policy chatbot

A prompt to write yours:

Write 3 LinkedIn headlines under 200 characters for a fresher who wants a [role] job.
My proven skills: [skills]. My best project: [one line].
Use words recruiters search for. No buzzwords like "passionate" or "ninja".

Expected output (yours will be similar, not the same): 3 short headlines in the "role | skills | proof" shape. Pick one, and make sure your About section and Featured section link the same project.

7. Where to look, and how to spot fake offers

Where to look:

1. Job portals such as Naukri, LinkedIn Jobs and Indeed, with the role words from section 2.
2. Internship portals such as Internshala for a first 2 to 6 month role.
3. Company career pages of IT services firms, product companies and GCCs (global capability centres) in your city.
4. Your own network: a short post about your project often brings more replies than 50 cold applications.

Fake offer warning signs:

• They ask you to pay a registration, training or "laptop security" fee. Real employers do not.
• The offer comes on WhatsApp or Telegram with no interview and a very high salary.
• They ask for your OTP, UPI PIN, Aadhaar or bank login.
• "Task-based" jobs: you get small payments for likes or ratings, then they ask you to "invest" to unlock more.

If something feels wrong, stop and report it at cybercrime.gov.in or call 1930.

8. Beginner mistakes and quick fixes

9. Try at home

Recap

Pick one role today, build one project this week, and apply when two projects are ready. Proof beats promises in every interview.

Previous chapter: AI Certifications Compared: AWS vs Microsoft vs Google, India 2026

Next chapter: 50 Generative AI Interview Questions and Answers (Beginner to Practitioner)

Ravindra Bagale, trainer: linkedin.com/in/ravindra-bagale. Neha, her degree, her skills and the job post are made up for learning. Job portals and reporting channels were checked on 19 October 2026. The Python output was produced on 19 October 2026. AI output changes every time, so your output will be similar, not the same.