Story
Ananya's Application Year
👩🎓 Ananya · Coimbatore · Age 17
Ananya wanted to study AI but didn't know whether to aim for IIT, IIIT, BITS, or apply abroad. She had no PhD parent guiding her. So she made her own roadmap — three flagship GitHub projects, a Kaggle bronze, one Medium article series, and a clear pitch.
Result: BITS Pilani CSE (Pilani campus) admit + IIIT-Hyderabad CSD admit + an unpaid summer research internship with an IIT-Madras professor. Total prep cost: ₹0 beyond JEE coaching. Roadmap and templates here.
Portfolio
The GitHub Portfolio That Gets Reviewed
College admission committees and internship recruiters skim — they don't read. Your GitHub must pass the 30-second test:
The 3-flagship rule: Pin three projects (not 30). Each should:
- Solve a real, named problem (not "ML demo").
- Have a README with: problem statement, dataset link, results table, deployed demo URL or video, lessons learned.
- Show one technical depth (e.g., "I implemented PagedAttention from scratch" or "I built the eval harness for 8 Indian languages").
Example flagship from this course:
- Class 11 L12 — Crop Disease Detection Web App (computer vision, deployed on Streamlit)
- Class 12 L01 — Marathi Tutor with QLoRA (LLM fine-tuning, Hindi/Marathi eval)
- Class 12 L08 — vLLM Deployment for 50K Users (production AI, k8s)
That's already three flagships, all built across this course. Polish the READMEs and pin them.
Indian Colleges
AI-Strong Indian Colleges (2026)
| Tier | Colleges | Entry Path | Notes |
|---|---|---|---|
| Tier 1 — IITs | IIT Bombay, Delhi, Madras, Kanpur, Kharagpur | JEE Advanced | Strong CS faculty; AI specialisation typically via electives + B.Tech project |
| Tier 1 — Specialised | IIIT Hyderabad (CSD/CLD), IIIT Delhi (CSAI), IISc UG (CS & Data Science) | JEE Main, UGEE (IIIT-H), KVPY/IISc admission test | India's top dedicated AI/CS programs |
| Tier 1 — Private | BITS Pilani (Pilani/Goa/Hyd), Plaksha, Ashoka | BITSAT, own tests | BITS CS at Pilani is on par with IITs in AI hiring |
| Tier 2 | NITs (Trichy, Surathkal, Warangal), VIT Vellore, IIIT Bangalore (PG only), Manipal | JEE Main, VITEEE | Solid CS, depends on student initiative |
| Specialised AI | IIT Hyderabad B.Tech AI, IIT Jodhpur AI & DS, IIT Patna AI | JEE Advanced | Newer AI-named programs at IITs |
Truth: Once you're at any tier-1 or strong tier-2 college, your portfolio + internships + GitHub matter more than your college name for AI roles. Colleges open the first door; your work keeps it open.
Global Colleges
If You're Considering Abroad
For undergrad AI/CS abroad, the typical strong options:
- USA: CMU SCS, MIT EECS, Stanford CS, UC Berkeley EECS, Princeton CS, UIUC CS, Georgia Tech CS, UMich CS. Plus need-blind for Indian students: Harvard, Yale, MIT, Princeton (limited).
- UK: Cambridge (Computer Science), Oxford (CS or Maths & CS), Imperial Computing, UCL CS, Edinburgh AI.
- Singapore: NUS CS, NTU CS — well-funded, strong AI faculty, Indian-friendly.
- Canada: University of Toronto (Geoffrey Hinton's home), University of Waterloo CS (great co-op), UBC CS.
- Europe: ETH Zurich CS, EPFL CS, TU Munich AI, KU Leuven AI.
Cost reality: US private undergrad ≈ ₹70 lakh/year all-in. UK ≈ ₹45 lakh/year. Singapore ≈ ₹25 lakh/year. Most need significant scholarship or family resources. Check need-based aid: MIT, Princeton, Yale offer it to international students; Stanford partial.
Pragmatic path: B.Tech in India (IIT/IIIT/BITS) → MS in USA/UK with research-funded admission. This is how most successful Indian AI researchers got started — and you graduate with portable income, not portable debt.
Internships
The Internship Hunt — Cold Email Template
Cold email is the most under-used tool by Indian students. Here is what worked for Ananya:
Subject: Class 12 student — your DDP paper inspired my IndicBERT pre-training run
Dear Prof. [Name],
I'm Ananya Krishnan, a Class 12 student in Coimbatore. After reading
your 2024 paper on efficient gradient compression for distributed
training, I implemented a small-scale reproduction on IndicBERT-base
across 8 GPUs and saw a 2.1x throughput improvement matching your
Table 3.
GitHub: github.com/ananyak/indicbert-ddp-repro
(Includes the modified all-reduce hook, training logs, and a write-up.)
I'd love to spend my summer (May-July 2026) helping with any
infrastructure or empirical-eval task in your lab — paid or unpaid,
remote or in-person. Even logging experiments would be a privilege.
Open to a 15-minute call any time that suits you.
Thank you for reading,
AnanyaWhy this works: Specific paper + concrete reproduction + GitHub link + flexibility on terms + zero-pressure ask. 1 in 8 emails like this got a positive reply for Ananya.
Other internship sources:
- Kaggle: A bronze medal is a real signal. Compete in Indic NLP / vision challenges.
- Open source: Contribute to AI4Bharat, Indic-NLP-Library, vLLM, transformers. A merged PR is a portfolio item.
- Smerc/RIA/IAS: Indian AI summer schools (often free for top applicants).
- Build-in-public: A working AI project with 100 real users beats most internships on a resume.
Career
The First 5 Years — Realistic AI Career Path
| Year | Role | Typical Compensation (India) | What to Build |
|---|---|---|---|
| Year 0 (UG) | Student + Intern | Stipend ₹15K–60K/month | Portfolio + 2 internships + 1 paper if possible |
| Year 1–2 | Junior ML Engineer / AI Engineer | ₹15–35 LPA (tier 1) or ₹8–18 LPA (tier 2) | Ship 2 production models; learn one specialisation |
| Year 3–5 | Senior ML / Research Engineer | ₹30–80 LPA | Lead a model surface end-to-end; mentor 1–2 juniors |
| Year 5+ | Staff / Principal / Founder / PhD | ₹60 LPA – ₹2 Cr+ or research/founder route | Architect or research direction; or start your own |
Top employers in India for AI roles (2026): Google India, Microsoft IDC, Amazon (AWS / Alexa), Adobe, Uber India, Ola Krutrim, Sarvam AI, Haptik, Glance, Razorpay, Flipkart, Swiggy, Zomato, plus US companies hiring remotely (Anthropic, OpenAI, NVIDIA, Cohere). Indian foundation-model startups (Sarvam, Krutrim, TWO AI) are the highest-leverage early-career bets right now.
Ananya's outcome: BITS Pilani CSE + IIIT-H CSD admits + IIT-Madras summer research. She picked BITS Pilani — closer family, strong AI alumni network, and the founder mindset she wants. Her GitHub now lists 5 flagship projects. She is on track to apply for an MS at CMU or Stanford in 4 years.
📝 Check Your Understanding (8 Questions)
1. What is the '3-flagship rule' for an AI portfolio?
a) Build at least 30 small projects to show breadth
b) Pin exactly three projects on GitHub, each solving a named real problem with a clear README (problem, dataset, results, demo URL, lessons), and each demonstrating one technical depth — admissions and recruiters skim in 30 seconds, so depth-on-three beats breadth-on-thirty
c) Have one flagship for each of the three FAANG companies
d) Have three flagships only if you are applying abroad
2. Which Indian colleges have dedicated UG AI/CS programs that rival the IITs?
a) Only the IITs offer real CS or AI programs
b) IIIT Hyderabad (CSD/CLD), IIIT Delhi (CSAI), IISc UG (CS & Data Science), and BITS Pilani CSE — these have strong AI faculty, dedicated curriculum, and place graduates into AI roles on par with the top IITs
c) Any private engineering college will do
d) Only colleges with 'AI' in their name count
3. Why does Ananya's pragmatic path recommend B.Tech in India then MS abroad?
a) Indian degrees are not respected internationally
b) A funded MS or PhD abroad (assistantship + tuition waiver) lets you study at top global universities without family debt; doing UG abroad costs ₹25-70 lakh/year out of pocket. Save the global move for graduate school where research funding is normal
c) It is the only way to get a US visa
d) MS is required to get any AI job
4. What makes a cold email to a researcher likely to get a reply?
a) Sending the same generic email to 500 professors
b) A specific paper reference + concrete reproduction or extension + GitHub link with working code + flexibility on terms (paid/unpaid, remote/onsite) + a low-pressure ask. Demonstrating you have already done unpaid work in their domain is the strongest possible signal
c) Mentioning your JEE rank in the first line
d) Attaching your full CV as a 5-page PDF
5. Why is contributing a merged PR to AI4Bharat or vLLM a strong portfolio item?
a) Open-source contributions count as paid work experience
b) A merged PR is independently verified evidence that maintainers (often senior engineers/researchers) judged your code production-worthy — a stronger signal than self-built projects which only you have evaluated
c) It is the only path to a software job
d) Open-source PRs guarantee an internship offer
6. What does Ananya's lesson say about college brand vs portfolio?
a) College brand always determines salary and role
b) College opens the first door (campus placement, alumni network); after that, portfolio + internships + GitHub determine your trajectory in AI roles. Strong work from a tier-2 college consistently beats weak work from a tier-1 college
c) Portfolio matters only at FAANG companies
d) Brand doesn't matter at all in India
7. Which Indian foundation-model startups are highlighted as high-leverage early-career bets in 2026?
a) Only multinational companies hire AI talent in India
b) Sarvam AI, Krutrim (Ola), TWO AI, and similar Indian foundation-model startups — small enough that you ship work that ships, large enough to have real GPU budgets, and building India-specific AI where domain knowledge gives an edge over global competitors
c) Government PSUs are the only stable option
d) Indian startups never hire freshers
8. What is the realistic compensation range for a Year 1-2 AI engineer at a tier-1 Indian campus in 2026?
a) ₹4-6 lakh per annum
b) Roughly ₹15-35 LPA at tier-1 campuses (IIT/IIIT/BITS), ₹8-18 LPA at tier-2 — actual numbers depend on company tier, internship conversions, and demonstrated AI specialisation; foundation-model startups and US-remote roles can pay higher
c) ₹2 crore per annum guaranteed
d) Compensation does not vary by college