Most data engineers already have their hands full with pipelines, data infrastructures, and reliable operations. Generative AI comes into play in addition to that with production service of models, management of retrieval systems, and construction of agent workflow systems which have to be as reliable as any other data pipeline. Infrastructure thinking is not as different as one would think, but the services have changed.
IIT Kharagpur Department of Computer Science and Engineering’s Executive Post Graduate Certificate in Generative AI & Agentic AI is tailor-made for filling this precise void. It lasts for 8 months, entirely live online, and covers learners from foundation models to deployed AI systems.
Where Data Engineering Skills Already Overlap
A data engineer's existing skill set maps onto this curriculum more directly than it might into a generic AI course, since much of the work is still about building reliable systems, managing infrastructure, and handling data at scale.
Experiences with building pipelines apply straightaway to creating RAG retrieval and reranking systems
Knowledge about API and systems design is assumed, not something that will be taught from scratch
Production-level infrastructure experience translates to working with LLMs in regard to latency and costs
Data pipeline reliability skills transfer to maintaining the stability of multi-agent workflows at each step
Comfort with structured data systems translates into vector database and unstructured data systems
Inside the Curriculum
The program follows a fixed progression, moving from core concepts to systems that run in production rather than staying at the prompting stage.
Foundation Models for GenAI and LLMs – transformer architecture and foundations models
Advanced Prompting & RAG – hybrid search and ranking
LLM fine-tuning and alignment - PEFT, LoRA, QLoRA
Multimodal and agentic AI – agent workflows planning and using tools
Deployment and optimization of AI applications – latency, cost, reliability
Do you aspire to learn more about Generative AI? Get started with the EPGC program on Generative AI & Agentic AI at IIT Kharagpur and acquire the industry-oriented skillset for the future. |
|---|
Hands-On Deliverables You'll Walk Away With
The program is structured around five deliverables rather than isolated assignments, so each module builds toward something that can actually run.
An enterprise retrieval RAG pipeline for evaluating retrieval accuracy
A fine-tuned open-source LLM using LoRA/QLoRA with an API deployment
A multi-agent system for performing tasks in multiple steps
A monitored generative AI API in a container for production use
A two-week capstone project by students in industry under faculty supervision
Fee Structure and Who Can Apply
The program's pricing is done with flexibility in mind, and there are also EMI options along with the complete payment method. Eligibility for the program depends not only on academic qualifications but also on the practical technical skills.
The cost of the program is ₹1,99,000 all-inclusive with taxes
You can reserve your seat through an advance deposit of ₹10,000
EMIs are available at ₹6,825 per month
Minimum 50 percent marks in graduation are mandatory for application
Students with other degrees may also apply with 2+ years of technology experience
In terms of hands-on experience, you must have the ability to write python functions, manipulate basic data structures, make use of API, and understand technical documentation
The program concludes with the certification being offered to you on campus, by the Programme Director and the Institute management
Steps to Enroll:
Enrolling follows four steps, start to finish.
Along with your application, submit all your details.
Then wait to get shortlisted and your provisional offer letter.
Make sure to get your slot confirmed with the submission of the necessary documentation.
Pay your remaining dues after the process of verification.
The IIT Kharagpur Advantage
The reputation that IIT Kharagpur has does not stem from marketing but stems from the research and academic prowess that IIT Kharagpur has had for several decades. The reputation is what informs the design of the program.
Founded in 1951, IIT Kharagpur holds the distinction of being India's very first IIT
NIRF 2025 places its 5th nationally in Engineering
Instruction is grounded in active research, with faculty regularly publishing at NeurIPS, ICML, ICCV, ACL, and IEEE
Every session is led by CSE department faculty directly, with no external or outsourced trainers involved
Completion is marked with a physical, on-campus ceremony rather than just a digital certificate
Conclusion
It's not the infrastructure for a data engineer which presents challenges when it comes to implementing generative AI, it's getting used to serving and managing systems which act different from what one normally finds in data pipelines, which require GPU-aware scaling, workflow calls of outside systems, and accurate retrieval system as the data evolves.
The whole course is designed on the very same idea, offering actual deployment of such systems as RAG pipelines, fine-tuned LLMs, and multi-agent workflows, instead of mere theoretical exercises. For an experienced data engineer who has already dealt with API design, systems architecture, and production infrastructure, this program is what he or she needs.
FAQs
How is building for Generative AI different from typical data engineering work?
Traditional data engineering focuses on moving, transforming, and storing data reliably at scale. Generative AI systems add a layer where the data engineer also needs to serve models, manage retrieval pipelines, and monitor non-deterministic outputs. This shift means working with new components like vector databases and model-serving infrastructure alongside familiar ETL work.
What changes when a data engineer starts deploying LLMs instead of standard data pipelines?
Standard pipeline monitoring tracks data quality and throughput against fixed expectations. LLM deployment adds monitoring for latency, cost per request, and output drift, since responses vary even for similar inputs. This program's deployment module and capstone project are built around exactly that kind of production monitoring.
Can data engineers apply Generative AI skills directly to their current work?
Yes, common applications include building retrieval pipelines for internal knowledge bases, serving fine-tuned models as internal APIs, and automating multi-step data workflows using agents. These use cases extend existing infrastructure work rather than replacing it. Most teams end up combining generative AI components with the data systems they already maintain.
What career roles can data engineers target after this program?
Typical next roles include Generative AI Engineer, MLOps Engineer, AI Infrastructure Engineer, and LLM Platform Engineer. These positions combine data engineering fundamentals with the ability to serve and scale generative AI systems. Several of these roles also carry a pay premium tied to the specialized skill set.
Do data engineers need prior experience with LLMs before starting this program?
No, the curriculum is sequenced to start from GenAI foundations before moving into fine-tuning, RAG, and agentic systems. Python skills, API experience, and comfort with technical documentation are enough to begin. Direct LLM experience is useful but not assumed going in.
How does Agentic AI extend what data engineers already know about workflow orchestration?
Agentic AI systems plan tasks, call external tools, and coordinate multi-step workflows instead of following a fixed pipeline. For a data engineer, this is closer to designing an orchestration layer with dynamic branching than to running a static ETL job. The program's multi-agent project applies this directly to task execution scenarios.
Is this program only useful for experienced data engineers, or can newer professionals benefit too?
The program assumes basic Python and API familiarity rather than years of experience specifically. Both early-career and experienced data engineers follow the same curriculum, though those with more hands-on infrastructure background may move through foundational sections faster. The eligibility criteria are based on educational background and, for other disciplines, prior technology experience.
How does a research-led faculty affect the quality of a Generative AI course?
Faculty who actively publish at venues like NeurIPS, ICML, and ACL tend to teach concepts as they currently stand in research, rather than repackaging tool tutorials. This matters in a field like generative AI, where techniques and serving methods change quickly. It also means course content is less likely to lag behind current deployment practices.
Will learning Generative AI make a data engineer's existing infrastructure skills less relevant?
No, most real-world generative AI systems still depend on solid data infrastructure underneath, whether that's data pipelines feeding a RAG system or monitoring feeding a deployed API. Generative AI skills add to that foundation rather than replacing it. The two are typically used together in production systems.
What makes a capstone project useful for a data engineer's portfolio specifically?
A capstone that goes from a business problem to a deployed, containerised system demonstrates more than model-building skill, it shows the ability to handle serving, monitoring, and scaling end to end. For data engineers, this fills a gap that many purely theoretical courses leave open. It also gives something concrete to walk through in interviews.