Top 10 Best Generative AI Courses in India (2026)

Coursera is the platform that most Indian learners already have an account on, and its generative AI catalogue is the widest in this comparison: professional certificates and specializations from Google, Microsoft, AWS, IBM, Vanderbilt and DeepLearning.AI, all sold through one Coursera Plus subscription or free to audit. This review covers the catalogue as a route — the two Coursera programmes with the deepest GenAI engineering content, DeepLearning.AI and IBM, are reviewed separately at #7 and #8.

Be clear about what the subscription buys. It is not a course — it is optionality: a beginner can start with Google Cloud's Introduction to Generative AI and Vanderbilt's Prompt Engineering specialization, a developer can move into the Microsoft AI & ML Engineering certificate or the AWS Generative AI Applications certificate, and a manager can stop at the Google AI or Generative AI Leader certificates. Nobody sequences that for you.

Coverage depends entirely on which programmes you string together. Across the big-brand certificates you can assemble Python and ML foundations, LLM fundamentals, prompt engineering at genuine depth (Vanderbilt), LLM APIs, introductory RAG with vector databases, agent building on Azure (Microsoft's certificate includes an agent-development course), Bedrock-based application building with guardrails (AWS), and responsible AI [VERIFY: current programme contents].

The catalogue's breadth is simultaneously its asset and its cost. Every programme is scoped to its sponsor's platform and audience, so a learner who wants portable GenAI engineering has to design the sequence, spot the overlaps and fill the gaps — production RAG evaluation, hands-on LoRA/QLoRA and agent reliability are thin everywhere except the DeepLearning.AI short courses.

Honest depth verdict: Excellent breadth and strong prompt-engineering and cloud-platform coverage; GenAI engineering depth is uneven and self-assembled. Hands-on fine-tuning, MCP, evaluation harnesses and LLMOps are limited outside the DeepLearning.AI library. Strong for a learner who knows what to pick; weak for one who needs to be told.

Fully self-paced video, readings, auto-graded quizzes and cloud-hosted labs. No cohort, no live IST session, no human reading your code. Forums exist and are patchy. The platform itself is polished, the mobile app is good, and audit access means you can try any course before you pay.

Completion is the honest problem: with no deadlines that bite and no mentor, most subscriptions become a monthly fee for good intentions. The learners who finish are the ones who already have a study habit.

Expect 3–6 guided labs per certificate — Bedrock console exercises, Azure agent builds, prompt-engineering assignments — well-specified but scaffolded, so the finished artefact looks like everyone else's.

What you get fewer of: an original, deployed system with retrieval metrics and an evaluation report. If your target role is titled *GenAI engineer*, you will need to build and document that yourself, outside the platform.

Free to audit; Coursera Plus is roughly ₹14,000 a year at list price in India, frequently discounted to ₹7,000–₹8,000 [VERIFY: current India pricing on the Coursera Plus page]. Individual certificates are also sold monthly. The subscription is the only way to get graded work and the certificate — and it keeps billing whether or not you log in.

Certification is the sponsor's certificate (Google, Microsoft, AWS, Vanderbilt) issued through Coursera. Duration runs 1–6 months depending on the programme and your pace [VERIFY].

Value is exceptional if you finish and self-assemble a coherent path; near zero if the subscription lapses in month three with two courses half-done, which is the common outcome.

There is no career-services layer — no interview preparation, no portfolio review, no partner network. Certificates are respected as evidence of initiative and largely ignored as a hiring credential on their own.

Coursera's own Job Skills Report is useful market context, not a placement mechanism. Treat any 'career outcomes' language on a certificate page as marketing until you can see a denominator.

If breadth and a recognisable certificate at subscription prices are the goal, this is the strongest catalogue available to an Indian learner. Just buy it knowingly: a library of courses, not a programme — you are the course designer.

DataCamp is the data-skills platform many Indian analysts already use for Python and SQL, and it has rebuilt its AI catalogue around three tracks: the Associate AI Engineer for Developers track (10 courses, about 29 hours: OpenAI API, prompt engineering, Hugging Face, LLMOps concepts, embeddings, Pinecone, LangChain and an introduction to MCP), the Developing AI Applications track and the Developing Large Language Models track (transformers in PyTorch, local Llama, RLHF) [VERIFY: current track contents]. Its defining value is every lesson is a coding exercise in the browser, graded instantly.

That format matters for a specific reader: the beginner or analyst who learns by typing, not by watching, and who cannot justify a lakh-rupee programme to test whether GenAI engineering is for them. For that reader the ₹600-a-month Premium plan — with the Associate AI Engineer certification included — is the cheapest serious entry point on this list.

Modules span Python for AI, the OpenAI and Responses APIs, prompt engineering, Hugging Face models, embeddings and semantic search, Pinecone vector databases, LangChain application building, LLMOps concepts, software-engineering principles, an introduction to MCP, and — in the LLM track — transformer architecture, fine-tuning Llama-family models and RLHF [VERIFY: current module list].

The organisation is bite-sized by design: four-hour courses of short videos and exercises. The constraint is depth per topic — a four-hour course on RAG cannot reach chunking strategy, hybrid search, re-ranking or a retrieval evaluation harness, and agent frameworks, guardrails and deployment get an hour each at most.

Honest depth verdict: Wide, current and hands-on at the introductory level; shallow past it. Prompting, LLM APIs, embeddings, introductory RAG and LLMOps vocabulary are handled well, and MCP is unusually present for a platform this cheap. Production retrieval evaluation, multi-agent reliability, guardrails and real deployment are light to absent [VERIFY].

Fully self-paced, in-browser. No cohort, no live session, no mentor. Doubt resolution is a community forum and an AI assistant inside the exercise — fine for a syntax error, useless for *why is my retrieval returning confident nonsense*.

Where DataCamp beats every other self-paced option here is friction: nothing to install, instant feedback, and daily-streak mechanics that get beginners past the first fortnight. Working developers often find the pace slow and the hand-holding heavy.

Expect 4–8 short guided projects across the tracks — a chatbot on the OpenAI API, a semantic-search demo, a LangChain application — plus a 4-hour practical exam in the certification. All are scaffolded inside DataCamp's environment.

None are deployed and none carry retrieval metrics. Treat them as practice, then rebuild one of them outside the platform with deployment, monitoring and an evaluation report if you want it to survive a technical interview.

Free tier (first chapter of every course); Premium is listed at roughly ₹600 a month billed annually in India — about ₹7,200 a year [VERIFY on the pricing page]. Monthly billing is materially more expensive. The Associate AI Engineer for Developers certification — a two-hour timed exam plus a four-hour practical — is included in Premium.

Certification is DataCamp-issued. Track durations are 1–4 months at a few hours a week [VERIFY].

This is a practice-weighted purchase: you are paying for volume of graded exercises and a low barrier to starting. That is a legitimate thing to buy. It is a bad deal only if you believed you were buying engineering capability or a credential that clears HR on its own.

There is no career-services team, no interview preparation and no partner network — the certification page and a professional profile are the whole career layer.

Ask nothing about placements, because there is nothing to ask. Budget a second, deeper investment if a GenAI-titled role is the goal.

The best choice on this list when you want to start coding GenAI today for the price of a streaming subscription — and a poor one if you're buying for engineering depth, mentorship or a credential. Know which of the two you're doing before the annual plan renews.

Great Learning is a long-running, operationally mature provider with McCombs (UT Austin) and Great Lakes branding on its applied AI portfolio — the partnership is confirmed on UT Austin McCombs' own site, and the current GenAI variant is listed as the PG Program in AI Agents & Generative AI for Business Applications alongside the longer PG-AIML programme (confirm the current partner and program name). Its applied GenAI programs are built around live weekend mentor sessions in IST — the signature of the format.

That scheduling decision defines who it suits. A consultant, a BFSI analyst or an IT-services lead who cannot reliably protect weekday evenings but can protect a Saturday morning will finish here and stall elsewhere. The alumni base is among the largest in the country.

Coverage runs Python foundations, ML and deep-learning essentials, LLM fundamentals, prompt engineering, LLM APIs, RAG with vector databases, introductory fine-tuning, introductory agents, multi-modal use cases, responsible AI, and business applications (confirm the current module list).

Sequencing is one of the better-considered on this list — each block genuinely prepares the next, and business framing is woven in rather than bolted on, which suits domain professionals.

Honest depth verdict: Solid, well-sequenced applied GenAI; not deep on production RAG evaluation, hands-on fine-tuning or agent reliability, and LLMOps is light. Refresh cadence is better than most university-affiliated programs and still behind specialists (confirm the last curriculum revision).

The model is recorded core content plus live weekend IST mentor sessions, and the mentors are practising professionals rather than career lecturers. Sessions are discussion-oriented — bring a problem and you get a conversation, not a slide deck.

Learner-support operations are strong: deadlines, nudges, program managers who chase you. This is a large part of why cohorts here actually finish, and it deserves as much weight as the syllabus in your decision.

4–6 GenAI projects with mentor feedback plus a capstone — and the feedback loop is one of the better ones in this price band, which is where most premium programs quietly fail.

Projects are applied and well-scoped, but few are deployment-grade. You will finish able to explain and demonstrate; you will need extra independent work to be able to say *this is running in production and here is how I measure it*.

₹1–3.5L by variant (see the program page), with EMI widely available. Duration 4–12 months. Certification carries UT Austin / Great Lakes branding — confirm the exact wording on the certificate before paying, and note that McCombs describes it as a certificate of completion from Texas Executive Education, not a degree.

Where the value sits: the mentor-led format and the brand, plus an operations team that gets people over the line. GenAI depth per rupee is moderate against specialist programs — a real trade, not a flaw.

Ask about API credits and whether the capstone requires paid inference; unbudgeted API spend is a common surprise across this whole category.

Resume review, mock interviews, a job board and a large alumni network. Assistance, not a guarantee.

Because the program attracts many domain professionals, a lot of good outcomes here are internal moves rather than new-employer placements. Ask for that split, and for the GenAI-role breakdown.

One of the most reliably completable premium programs for a working Indian professional. Choose it for structure, mentorship and brand — not for frontier generative AI depth.

Intellipaat is a large Indian EdTech provider offering generative AI certifications in collaboration with IIT institutes — the current program page advertises a certificate "from IIT & Microsoft" (confirm the current affiliation and which IIT). It sits deliberately between budget platforms and premium university programs: an institutional tag plus reasonable applied depth at a third to a half of premium pricing.

For a cost-conscious buyer who still needs something recognisable on the CV, that positioning is genuinely useful — provided you go in knowing that the support experience is self-driven.

Modules typically cover Python, ML essentials, deep learning, NLP, LLM fundamentals, prompt engineering, LLM APIs, RAG with vector databases, introductory fine-tuning, introductory agents with LangChain, multi-modal content, and cloud deployment components (confirm the current module list).

The deployment exposure is the pleasant surprise — several higher-priced competitors on this list offer less of it. If you want to finish able to put something on a cloud instance rather than only in a notebook, that matters.

Honest depth verdict: Broader and more deployment-aware than most mid-tier GenAI programs; agentic and evaluation depth is moderate; quality varies noticeably by module and instructor. Ask which instructor teaches your batch, and request a module-level syllabus with revision dates before paying.

A hybrid model: self-paced content plus live instructor-led sessions, with 24/7 support claims that you should test during pre-sales rather than after paying. Ask a technical question over chat at 10pm and see what comes back.

Larger cohorts dilute mentor attention, and consistency across modules is the most common learner complaint in this band. Recordings are available and the LMS is functional rather than polished.

4–8 projects with industry-scenario framing plus a capstone, and some genuine deployment exposure.

Review depth varies and is not consistently code-level. Where feedback is thin, the fix is to publish the project on GitHub with your own written architecture rationale and evaluation notes — that is what an interviewer reads anyway.

₹60K–₹2L by variant (see the official page), EMI available, with frequent discounting — negotiate, and get every inclusion (mentor hours, API credits, career services, certificate wording) confirmed in writing. Duration 4–9 months.

Certification is IIT-affiliated — clarify precisely what the affiliation covers, because 'affiliated' spans everything from joint curriculum design to a two-day campus immersion.

Value is good specifically for the credential-plus-breadth combination. It is not the best depth per rupee on this list, and it does not pretend to be.

Job assistance, resume preparation and mock interviews. Verify the currency of any partner list — hiring-partner logos age faster than websites update.

Ask for the GenAI-role breakdown and the denominator behind any placement percentage (confirm the current published outcome data).

A sensible middle path for breadth plus an institutional tag without premium pricing — with the standing caveat that you must drive your own support experience to get full value from it.

Simplilearn is a certification-led global platform, with Purdue University and IBM collaboration on its applied generative AI portfolio — the Purdue tie-up was announced in 2024, but on the check date the live Applied Generative AI Specialization page named Michigan Engineering Professional Education and Microsoft Azure as partners (confirm the current partners and program name before quoting either). Its genuine advantage is not curriculum depth — it is corporate legitimacy.

It is one of the most commonly employer-reimbursed platforms in India, and its credentials are familiar to HR and L&D teams. If your learning budget flows through a procurement process, that familiarity is worth more than any syllabus comparison.

Coverage typically includes GenAI fundamentals, prompt engineering, LLM APIs, introductory RAG, GenAI for business workflows, introductory fine-tuning concepts, responsible AI, and a capstone (confirm the current module list).

The framing is workflow- and business-first, which suits managers, analysts and consultants who need to scope and evaluate GenAI work rather than build it.

Honest depth verdict: Broad and industry-oriented but moderate in depth — optimised for certification completion rather than engineering rigour. Agents, MCP, production RAG and LLMOps are not meaningful components. Do not buy this expecting to leave able to build and defend a production retrieval system.

Delivery is predominantly self-paced core content plus live 'masterclasses' — an important distinction, because marketing in this category routinely implies fully live instruction. Ask how many hours are genuinely live, and with whom.

Support is forum- and ticket-based with limited personal mentorship. Progress tracking is good. Completion therefore depends heavily on your own discipline — which is precisely why employer-mandated deadlines improve outcomes here.

3–6 projects with industry framing plus a capstone — structured, clearly briefed, and largely guided, with limited independent design and little real code review.

The honest read: these demonstrate exposure, not engineering judgement. They are useful evidence for an internal mobility case and weak evidence in a technical GenAI interview.

₹1–2.5L (see the official page — the 16-week Applied GenAI variant listed there sits inside that band), EMI available, with frequent promotional pricing — never pay a list price on a first call. Duration 4–11 months.

Certification carries university and vendor branding (Purdue / IBM on older variants; Michigan Engineering / Microsoft on the current page), which is the most reimbursement-friendly combination on this list.

The single most useful sentence about this option: strong value when employer-funded, moderate value when self-funded. Decide which you are before comparing it with anything else here.

Career services and a job board, oriented towards enterprise and services roles rather than product-company engineering pipelines.

Because much of the audience is already employed, a large share of good outcomes are internal role changes. Ask for that split rather than a headline percentage.

Excellent if your employer is paying and credentials matter internally; mediocre value if you're self-funding for engineering capability. That sentence should decide it for almost every reader.

Andrew Ng's programs are the global reference standard, and nothing else on this list comes close on conceptual clarity. *Generative AI with LLMs* (with AWS) gives the clearest available grounding in how LLMs are trained, adapted and evaluated, and the short-course library covers prompting, LangChain, RAG, vector databases, fine-tuning, evaluation, agents in LangGraph and MCP, each with hands-on labs.

Position it correctly and it is unbeatable: a foundation layer plus a supplement library, not a complete career program. Position it as a career program and you will be disappointed for reasons that have nothing to do with quality.

The flagship course covers LLM architecture and training, prompting, PEFT and LoRA, RLHF concepts, evaluation, and deployment considerations. The short-course library extends into RAG, embeddings, agents, multi-modal work, evaluation and LLMOps (confirm the current catalogue at deeplearning.ai/courses). If you want ML foundations first, the Machine Learning Specialization and Deep Learning Specialization are the canonical free-to-audit prerequisites.

The teaching on fine-tuning concepts and evaluation is better than in any paid Indian program on this list. If you want to genuinely understand why LoRA works rather than copy a training script, start here regardless of what else you buy.

Honest depth verdict: Unmatched clarity on foundations and fine-tuning concepts; deliberately fragmented in delivery. Agents and LLMOps material is spread across many short courses with no integration project, no Indian context and no production pipeline. You get excellent parts and must assemble the whole yourself.

Fully self-paced, with world-class production quality and excellent labs. There are no live sessions, no mentors, no code review and no cohort — forum support only.

This is simultaneously the format's strength and its fatal weakness. Self-paced completion rates across the MOOC category are famously low — the MIT analysis of six years of edX data in *Science* found about 3% of enrolments completing — and no amount of production polish changes that. If you have abandoned two self-paced courses already, that is data about you, not about the course.

The labs are high-quality and carefully scaffolded — they teach exceptionally well and demonstrate very little to a recruiter, because everyone's output looks identical.

To convert this into employability you must build separate portfolio projects of your own design: a RAG system on your own corpus with retrieval metrics, a fine-tune with a benchmark, an agent with failure handling. The course gives you the understanding to do that; it does not do it for you.

Free to audit; roughly ₹3,000–₹4,000 per month for a Coursera subscription if you want certificates and graded labs (confirm the current pricing on the Coursera Plus page). Realistic duration 2–6 months.

Unmatched value per rupee — with two cautions. Subscription creep is real: a six-month drift costs more than a budget Indian program. And API keys and any GPU time are self-funded, which surprises learners mid-way through the fine-tuning material.

Certificates are Coursera certificates. They are respected as evidence of self-motivation and effectively worthless as a hiring credential on their own.

None — and the platform is honest about that. There is no resume review, no interview preparation, no job board relevant to Indian GenAI hiring.

That honesty is worth noting: several paid programs on this list market career support that, in practice, delivers less than a determined learner achieves with LinkedIn, referrals and a good GitHub profile.

The best generative AI foundations available anywhere, and an incomplete answer to the question this page exists to answer. Pair it with your own structure, your own deadlines and your own projects — or pair it with a paid program and get more out of both.

A structured, applied certificate aimed squarely at producing practising GenAI engineers with widely used tooling — the official Coursera listing shows a 16-course sequence at roughly six months on 6 hours a week, with the LLM-engineering core also sold separately as a specialization. Where DeepLearning.AI explains, this implements — more notebooks, more integration, less theory (confirm the current program structure).

Two things make it worth ranking above several paid Indian programs on a value basis: it is far cheaper than any premium Indian option, and the IBM name registers in enterprise and IT-services contexts where recruiters recognise vendor credentials.

Modules typically cover Python for GenAI, transformer fundamentals, prompt engineering, LLM APIs, Hugging Face Transformers, LangChain, RAG with vector databases, introductory fine-tuning, model deployment basics, and a capstone (confirm the current module list on the certificate page).

The RAG and Hugging Face material is the strongest part — genuinely hands-on, with enough repetition to build muscle memory rather than a single demo.

Honest depth verdict: Strong applied breadth for the price, moderate theoretical depth, agents and MCP limited. LLMOps and production deployment are *touched* rather than taught — you will meet Docker and an endpoint, not observability, cost routing or prompt versioning.

Fully self-paced, with hands-on labs in cloud notebook environments that remove setup friction — a real advantage for learners on modest laptops or constrained bandwidth.

There are no live sessions, no mentors and no code review, so the same MOOC completion risk applies. Module sequencing and lab infrastructure are good enough that motivated learners do finish.

5–8 guided labs and applied projects plus a capstone — more build-oriented than typical MOOC assignments, and still guided.

To be portfolio-defensible, extend them into original work: swap in your own corpus, add retrieval metrics, deploy the capstone somewhere real, and write the architecture rationale yourself. Interviewers recognise course capstones instantly.

Free to audit; roughly ₹3,000–₹4,000 per month via Coursera subscription for graded work and the certificate (see Coursera Plus). Duration 3–6 months at a working professional's pace.

Certification is an IBM professional certificate — recognisable, vendor-backed, and no substitute for a portfolio.

Excellent value per rupee, with the same subscription-creep caution as any monthly plan: set a target finish date before you subscribe, and budget separately for API usage.

None claimed — no resume support, no mock interviews, no job board relevant to Indian GenAI roles.

In exchange you pay roughly one-fiftieth of a premium program. For a self-directed learner already in a job, that is often the better trade.

The strongest ₹0–₹5,000 option for someone who already codes, and the best applied-practice value in this comparison. Treat it as practice plus a recognisable name — not as a career program.

GUVI is an IIT-Madras-incubated platform (now part of HCL — see IIT Madras' own note on GUVI) whose defining strength is language accessibility: instruction in Tamil, Hindi, Telugu, Kannada and English, with dedicated Tamil and Telugu generative AI course pages (confirm the current language coverage). That single decision changes who can learn generative AI in India.

It deserves emphasis, because the rest of this list quietly assumes fluent technical English. For a capable diploma or B.Sc. graduate in a Tier-3 town, the barrier has been the language of instruction, not the difficulty of the material. The platform is mobile-first, bandwidth-conscious and priced for Tier-2/3 affordability.

Coverage typically includes Python, ML essentials, GenAI fundamentals, prompt engineering, LLM APIs, introductory RAG, and applied GenAI tooling (confirm the current module list).

Read the scope honestly: this is an entry and consolidation platform. It gets you from zero to competent user and junior builder in your own language — a large, genuinely valuable distance to travel.

Honest depth verdict: Solid foundational-to-intermediate GenAI coverage; minimal fine-tuning, agent frameworks, evaluation and LLMOps. Not a route to GenAI engineering on its own, and it does not claim to be.

Live and recorded sessions in regional languages, active regional communities, code playgrounds in the browser, and a platform designed for phones and patchy bandwidth rather than office broadband.

Regionally organised support meaningfully improves engagement for vernacular learners — a peer group that debugs in your language is worth more than a better-produced video in a language you translate as you watch.

2–4 GenAI projects at entry-to-intermediate level, with a capstone in some variants.

Enough to demonstrate foundational competence for entry roles and internships; not sufficient alone for competitive GenAI engineering positions. Plan a second, deeper investment once English technical content becomes comfortable.

₹10,000–₹80,000 by tier, with EMI available (see the course catalogue and the AI & ML program page). Duration 3–9 months.

Certification is an IIT-Madras-incubated platform certificate — 'incubated' means the company emerged from the institute's ecosystem, not that IIT-M awards or teaches the program. Say that plainly to yourself before you value the tag.

Very strong value within its band, especially where the honest alternative is no accessible option at all.

Regional placement support with genuine strength for Tier-2/3 entry-level roles — service companies, startups and analyst positions in regional hubs.

Do not expect a product-company GenAI pipeline. Ask for the regional breakdown and role titles.

The right first step for a large, underserved group of Indian learners — and best followed by a deeper program once English technical content stops being a tax on your attention.

PW Skills is Physics Wallah's skilling arm, applying the group's affordability-first philosophy to technical education — the Data Science with Generative AI course page is the program reviewed here. The defining feature is simply price: a structured, community-supported program at a fraction of every other structured option on this list, delivered in Hindi-English with a very large learner community.

Frame it correctly and it is excellent. This is a low-cost starting investment, not a complete career program — and for a student, fresher or curious professional, starting cheaply is a rational first move.

Coverage typically includes Python, statistics, data analysis, SQL, machine learning, introductory deep learning, and a GenAI component spanning LLM basics, prompting, LLM APIs, introductory RAG and LangChain basics (confirm the current curriculum).

Note the shape carefully: the GenAI layer is a module inside a data science course, not the centre of gravity. The SQL and analysis content is genuinely useful and is also most of what you're buying.

Honest depth verdict: Reasonable coverage for the price, but entry-level GenAI depth — no meaningful fine-tuning, agent frameworks, MCP, evaluation or LLMOps. Judge it against ₹10,000, not against ₹1L programs, and it looks good.

Primarily recorded content with live doubt-clearing sessions and a very active community, on a mobile-friendly platform.

Support is community-heavy rather than mentor-heavy, so quality depends on how actively you engage — lurkers get much less than participants. Recorded-first delivery also raises dropout risk despite the low price, which is the quiet cost of cheap education.

2–4 entry-level GenAI projects with guided walkthroughs — good for initial confidence and a first GitHub presence.

Not sufficient for a competitive GenAI portfolio without independent extension. The most valuable thing you can do here is take the guided project and rebuild it on data nobody else in the cohort used.

₹5,000–₹30,000, with EMI on higher tiers (see the official page). Duration 4–8 months. Certification is a PW Skills certificate — treat it as a completion record, nothing more.

This is the lowest-risk structured entry point in Indian GenAI education. The worst realistic outcome is that you spend ₹10,000 and learn that you don't enjoy the work — which is cheap, useful information.

Budget separately for API usage on your own projects; even small experiments cost something.

A growing placement cell with an entry-level focus — internships, analyst roles and junior developer positions.

Expect nothing resembling a GenAI engineering pipeline, and treat any percentage without a denominator as marketing.

Probably the best first ₹10,000 you can spend on generative AI in India — bought with clear understanding that a second, deeper investment is needed before you're competitive for a GenAI role.