The idea of a program built almost entirely around hands-on projects to demonstrate how generative AI models are used in real applications was extremely appealing to me. I expected a practical, industry-oriented experience that would deepen my understanding of how to work with modern AI systems in production contexts. At the beginning, the theoretical introductory lessons were genuinely excellent. The explanations of core concepts such as large language models, speech technologies, RAG, and the overall ecosystem around IBM watsonx and Hugging Face were clear and engaging. These initial modules were the strongest part of the course and provided solid foundational knowledge. However, as the course progressed, many of my initial concerns unfortunately proved valid. A significant portion of the hands-on work revolves around using specific models through Hugging Face or similar platforms. The problem is that model interfaces change frequently, APIs evolve, and specific models quickly become outdated. Investing time in learning how to use a particular model configuration feels of limited long-term value when that model may be replaced or deprecated within months. In practice, the course gives a general idea of how to interact with LLMs in Python, but it focuses too heavily on implementation details that are not particularly transferable or durable. Some of the models covered are already far less relevant in today’s rapidly evolving generative AI landscape. From a professional standpoint, I did not feel that this significantly improved my readiness to work with state-of-the-art systems currently used in industry. Another weakness is the structure of the labs. Many exercises feel passive: you are often copying and pasting large chunks of code rather than actively designing or reasoning about solutions. This approach limits deep understanding and does not encourage critical thinking or independent problem solving. It becomes more of a guided replication exercise than true hands-on engineering practice. The development environment was also frustrating. It frequently required restarts and felt unstable, which disrupted the learning flow and added unnecessary friction to the experience. In my opinion, the course would benefit from being more focused. Out of the seven projects, three well-designed, deeper, and more challenging projects would have been more effective than seven relatively superficial ones. More emphasis on architectural thinking, abstraction, and transferable design principles would significantly improve its long-term value. Overall, while the introductory theory and exposure to tools like Gradio and some LangChain basics were useful, much of the practical component felt superficial and quickly outdated. I finished the course feeling that the time invested did not translate into proportionate professional growth.

This course is badly in need of updates and the worst course I've taken from IBM Coursera. Multiple projects have use AI APIs from 2020 and some are currently broken due to deprecated tools and code.

IBM lab kicked me out every time. In this module, never get to complete downloading torch module. Lab WENT OFFLINE! How am I gonna be able to finish anything here??????????????