Anupam Sobti teaches Web Development using AI at Plaksha University.
Source note: This is an edited interview adapted from a narrated video submitted to OpenAI. Watch the associated video in OpenAI Academy.
Intro
What happens when you hand a cohort of motivated undergraduates the fundamentals of modern web development—and invite them to infuse every step with AI? Dr. Anupam Sobti, assistant professor at Plaksha University, set out to answer this question with his course “Web Development Using AI” (now called AI Product Design). In this conversation, Dr. Sobti shares how he reimagined the undergraduate learning experience, why he believes AI is a capability (not just a skill), and what happens when students are encouraged to build complete systems from early years rather than wait to learn all the fundamentals before ever building a complete product.
The Interview
Q: Dr. Sobti, what inspired this course?
Sobti: The core motivation was to bridge the “relevance gap” I noticed in undergraduate engineering. Students learn the fundamentals—data structures, programming, linear algebra—but often struggle to see how these connect to building useful software. With the rise of large language models and their utility in software workflows, I saw an opportunity to bring AI directly into the web development process. My goal was to move from a bottoms-up, theory-first approach to a journey where students go from idea to prototype, experiencing the full stack and the power of AI along the way.
Q: Can you describe at what stage of their undergraduate degrees did students take the course?
Sobti: We had a selective process—about 25% of applicants were accepted, from 2nd and 3rd year undergraduates who showed motivation and a clear vision for what they wanted to build. The only formal prerequisite was Python programming, which they had covered in their first semester. From there, we built up: students learned everything from HTML, CSS, and JavaScript to backend APIs, databases, and security best practices.
Q: Can you walk us through the workflow students followed, especially in terms of integrating AI?
Sobti: Every student started by proposing a project idea, which needed to include basic web development, LLM integration, and some form of Retrieval-Augmented Generation (RAG) or agentic pipeline. We taught them to use FastAPI for backend endpoints, and to prompt LLMs for generating everything from HTML and CSS to backend logic. Tools like Cursor, Lovable, and Bolt helped with rapid prototyping, while API endpoints provided more control and reproducibility—especially for database interactions. Later, we introduced agentic workflows and custom tool development, including crew AI for orchestrating multiple agents.
Q: What are some standout student projects that emerged from the course?
Sobti: The creativity was remarkable. One project, Bugzer, uses an LLM to automate bug-finding in apps—saving developers from manually checking dozens of pages after each update. Another, Luma, was inspired by a student’s experience with their father recommending books. Luma connects local libraries and readers, suggesting books based on mood and interest via an AI chatbot. Then there’s CourseVibe, an AI-enhanced course review and recommendation platform for Indian college students. It uses validation questions and reasoning models to ensure reviews are genuine, filtering out unreliable feedback before it’s published.
Q: How did students experience the learning process, especially when working with AI tools?
Sobti: Many students described “aha moments” when they realized AI wasn’t just an assistant but a capability that could be woven into any project. For example, one student unlocked new project functionality by enabling memory in their agent, allowing it to remember past interactions. Another shifted from using GPT in the browser to hitting the API directly, providing full codebase context and targeted fixes—this was a breakthrough in understanding the potential of AI. Across the board, students found that AI sped up learning and iteration, making it possible to move from concept to working prototype much faster than traditional methods. I also found students more interested in the core areas like data structures, databases, etc. once they were able to witness how these technologies made a significant difference in their products.
Q: What’s the broader impact you hope to see from this approach?
Sobti: I want students to realize that software in a human-enabler and with AI, we can broaden how much can be enabled for humanity. I want the students to be able to witness product journeys and build curiosity about fundamental computer science as well as advanced AI capabilities eventually integrating it into their workflows. I hope they see AI not as a niche skill, but as a capability that can be integrated into any domain. The speed of iteration and the ability to test, refine, and deploy ideas rapidly is transformative. Ultimately, I hope this approach encourages students to see connections across subjects and to approach problems with creativity and confidence.
What Stands Out
Core idea: AI is framed as a capability to be integrated throughout the software development process rather than a standalone skill.
Classroom design: The course centers on rapid prototyping from idea to working web application, combining frontend, backend, and AI workflows. The course helps students take a top-down approach to learning how systems work.
Student impact: Learners created projects addressing real-world problems, such as automated bug detection and personalized recommendations, while experiencing accelerated learning.
Transferable lesson: Embedding AI tools in project-based learning can deepen understanding and foster creativity by enabling faster iteration and practical application.
Bio
Dr. Anupam Sobti is an assistant professor at Plaksha University. He completed his postdoctoral research at Microsoft Research India and earned his PhD at IIT Delhi with the assistive technology group. His current research focuses on geospatial and embedded perception for agri-urban sensing.
What I find most interesting is that AI changes what "learning to code" actually means.
If AI can generate syntax on demand, then the lasting skill becomes understanding the problem, evaluating the solution, spotting what's wrong, and knowing how to improve it.
In that sense, AI doesn't reduce the need to learn web development—it shifts the emphasis from remembering code to reasoning about systems.
The strongest developers may not be the ones who type the most code, but the ones who understand it deeply enough to question, adapt, and extend what AI produces.