How to get hired in AI/ML without experience in India (2026)
The honest answer: stop applying. Start shipping.
Every fresher we've placed in 2026 has the same pattern. They didn't send 200 resumes and pray. They built three things, made them public, and one recruiter found them. Here's how to copy that.
1. The resume route is broken — accept it.
In 2025, India lost 244,851 tech jobs. ATS bots now reject 75% of applications before any human reads them. Even the ones that survive get 7.4 seconds of attention. With ChatGPT making every CV look identical, recruiters have stopped trusting them as a skill signal and pivoted to verifiable proof — code, demos, scores from trusted graders.
If your only job-hunt tool is a CV, you're competing in a market that doesn't believe CVs anymore.
2. Build three things — small, public, real.
You don't need a moon-shot. You need three projects that prove three different skills, each with a working demo URL. Recommended for an AI/ML fresher:
- Project 1 (week 1–2): a content recommender. Uses scikit-learn
+ Streamlit. Recommends Bollywood movies given a user's last 5. GitHub + a free Streamlit demo URL. Done.
- Project 2 (week 3–6): a fine-tuned LLM. Take a 7B model,
fine-tune on a niche corpus (Hindi customer-support tickets work great), evaluate intent-classification F1, and write a 500-word README explaining the trade-offs. Hugging Face + GitHub.
- Project 3 (week 7–10): a production RAG. Vector DB (pgvector
or Pinecone free tier), LangChain, a real document set (your college's prospectus, India's RBI circulars, anything specific). Deploy to Vercel. Show p99 latency in the README.
After three months you have a proof stack that beats most 2-year Bangalore engineer's CVs.
3. Get every project graded by something verifiable.
The reason recruiters discount projects is they can't tell who actually wrote them. ChatGPT can ship a Streamlit app in a weekend. The fix: a calibrated AI Score — a third-party rubric that recruiters can trust because it's the same scale across thousands of candidates.
ZT runs Claude against every submission with a fixed rubric (code quality, problem-solving, real-world fit). Score is 0–100, public, verifiable from your profile URL. The exact same rubric that gave a fresher a 78 yesterday gives a Google L3 a 92 today. That comparison is the value.
If you're not on ZT, the next-best move is open-source contributions that get merged into a meaningful repo (≥1k stars). PR description + review thread is the closest thing to a public score.
4. Be findable. Don't apply.
This is the inversion most fresher candidates miss. If you have a public score and a public portfolio, the search direction reverses. Recruiters search the talent pool by skill + score. Companies on ZT are doing this every day — filter by Python ≥ 80, fine-tuning ≥ 75, based in Hyderabad, available now. Three clicks.
To rank in their searches, you need:
- A public profile URL with your real name in the slug
(zerotheory.live/u/your-name, github.com/your-name).
- A clear specialisation in your bio — “Computer Vision
engineer” not “ML enthusiast.” Recruiters search on the specialisation phrase.
- Three pinned projects with working demo URLs at the top of
your profile. Not screenshots. Live URLs.
- A score of 70+ across at least one specialisation. Below 70 you
look unconfirmed. Above 88 you'll get poached at offer time.
The 72–88 range is the sweet spot for fresher hiring.
5. The first reach-out is from them, not you.
When this works, you don't apply. A recruiter messages you on LinkedIn or directly through a platform like ZT, says “saw your fine-tuning project, are you open to a 30-min call,” and you've already won.
What you say back: “Yes. Tell me about the team and the role. What's a hard problem they've solved this quarter?”
The phrasing matters. You're not desperate. You're asking whether they're worth your time.
6. Salary range is data, not negotiation.
For 0–2 year AI/ML engineers in India in 2026:
| Tier | Range | Where |
|---|---|---|
| Entry | ₹6–10 LPA | Startups outside Bangalore, agencies |
| Standard | ₹10–18 LPA | Bangalore / Hyderabad mid-market |
| Premium | ₹18–35 LPA | Razorpay, Flipkart, Swiggy, unicorns |
| Top | ₹35–60 LPA + ESOPs | FAANG India, OpenAI, Anthropic India teams |
Don't accept entry-tier if you have an AI Score above 80. The market genuinely pays. Use the score as your anchor.
What to do this week
- Sign up at zerotheory.live (free Week 1, no card).
- Pick the Python & AI/ML Fundamentals track if you've never
shipped, or the LLM track if you've already done a Coursera-style course.
- Ship Project 1 by Sunday. Submit it for AI grading.
- Make your profile public. Put the score on your LinkedIn header.
- Wait. Don't apply. Watch what happens.
If it doesn't work, email me. I read every email — kundankhatri@zerotheory.live.
— Kundan, founder of ZT