Build Production-Ready AI Agents and Multi-Agent Systems from JHU

If you are interested in AI and have a basic understanding of Python or programming, I highly recommend this program. It provides a strong foundation in Agentic AI with practical examples, hands-on learning, and excellent instructor support. The knowledge gained can be applied immediately to real-world projects and career growth.

Read More

I am a physician by training, and much of my professional experience has been at the point where healthcare systems meet real people: patients, families, junior medical staff, and the operational realities of hospitals. Over the years, I began noticing a recurring problem. Healthcare does not necessarily fail because information is unavailable. It can fail because the right information, in the right context, does not reach the right person at the right moment. That observation changed the questions I was asking. I became particularly interested in informed consent. I have seen consent administered countless times in both elective and emergency settings. The process can vary considerably depending on the clinician, the time available, and the circumstances. A form may be correctly completed, signed, and filed, while the patient may still not fully understand the procedure, its alternatives, its material risks, or what the decision actually means. In other words, the documentation can be complete while the communication remains incomplete. That led me to begin developing the concept of ConsentPod: a bedside system intended to transform consent from a document-signing event into a process of education, communication, comprehension, and shared decision-making. But I quickly encountered a limitation in my own thinking. Having an idea for an AI-enabled healthcare product was not enough. I needed to understand what actually happens beneath an AI system if I wanted to build something responsibly. I did not want to become merely a better user of ChatGPT. I wanted to understand agents. Those questions motivated me to undertake the Johns Hopkins Agentic AI program. I entered the program as a doctor trying to understand technology. I came out thinking much more like a systems designer. The most valuable shift was conceptual. I began to move from thinking about AI as a model to thinking about AI as a system operating within an environment. That distinction became particularly meaningful when I applied it to healthcare. One concept I kept returning to was context. An AI system may be able to access enormous amounts of information, but access to information is not the same as understanding. I increasingly believe the future of AI is not about making the context window larger, but about making the context better. This connected many of the technical concepts from the program, including retrieval-augmented generation, tool use, orchestration, memory and state, evaluation, and human-in-the-loop systems, to problems I had already encountered clinically. The program therefore gave me something more valuable than a collection of technologies. It gave me a framework for decomposing complex human problems into intelligent, evaluable workflows without losing sight of the human decision at the center. The clearest application has been my continued development of ConsentPod. Agentic AI gave me a way to think about how such a workflow could be constructed. Rather than building one general-purpose chatbot, I began decomposing the problem into responsibilities: explaining information in the patient’s preferred language, retrieving authoritative clinical and institutional information, checking comprehension, verifying that workflow requirements have been addressed, documenting the interaction, and escalating situations that require human attention. The physician would still make the clinical judgment. The patient would still make the decision. The AI would not become the doctor. It would become an orchestration, communication, and context layer around the decision. That distinction has become central to how I think about healthcare AI. My most obvious challenge was that I came to the program as a physician rather than as a computer scientist. But the deeper challenge was learning how to translate a clinical problem into a computational problem without stripping away the human complexity that made the problem important in the first place. Healthcare is not a clean software environment. It contains incomplete information, interruptions, uncertainty, hierarchy, time pressure, emotional vulnerability, legal accountability, and decisions whose consequences can be irreversible. My clinical background became an advantage here. Whenever I encountered a technical concept, I tried to translate it back to the bedside. That way of translating between technical architecture and clinical reality helped me overcome the initial learning curve. More importantly, it changed the way I approach technology. I no longer begin with where can I put AI. I increasingly begin with what is the human problem, what decision is being made, what context is missing, what workflow surrounds that decision, and where an intelligent system could responsibly improve it. For me, the greatest value of the Johns Hopkins Agentic AI program is not the certificate. It is the change in the way I think. I entered with a collection of healthcare problems and a growing fascination with artificial intelligence. I leave with a more structured way of thinking about how intelligent systems might participate in solving those problems. Most importantly, I have begun to see AI not as a destination, but as infrastructure for solving human problems. That is the perspective I hope to carry forward as I continue working at the intersection of medicine, technology, and healthcare delivery.

Read More

When I was offered this course, I was told that I did not need to know Python. I believe that was a fundamental mistake because, in order to successfully complete the hands-on labs, it is essential to have a solid understanding of Python. While it is true that you can complete the assignments, since the labs and projects often require writing only a few lines of code, I personally believe that many students rely heavily on artificial intelligence to finish the tasks without developing a true understanding of Python. The level of Python required for this course is actually quite advanced, as the course is much more development-oriented than focused purely on artificial intelligence itself. Regarding the instructors, I must honestly say that I am not completely satisfied. I have no doubt that they are highly professional and extremely competent in their respective fields. However, I found that most of the videos move too quickly, often giving the impression that the speakers are reading from a script. The explanations also tend to be somewhat superficial, and in many cases some concepts are not fully explained. I was left with the feeling that this course is about 90% Great Learning and only 10% Johns Hopkins University. At the beginning of the course, several instructors are introduced. However, in practice, I felt that William Roncal and Dr. Shelby were the ones who played the main role throughout the program. Unfortunately, neither the professor who worked for the FBI nor the military expert from the Gulf War contributed much, if anything, to the course. From that perspective, I felt somewhat disappointed. To be fair, I do not know whether they contributed significantly behind the scenes in ways that were not visible to students. If that is the case, then my opinion may be biased. Nevertheless, this is my honest feedback. Regarding the hands-on labs, Professor Hassam is a wonderful person with an excellent attitude and a genuine willingness to teach. However, I sometimes felt there was a slight lack of coordination or preparation regarding the material covered during the sessions. At times, I had the impression that Professor Hassam was reviewing the code for the first time during class rather than beforehand. That said, this may simply be my own mistaken perception. Overall, Hassam deserves a perfect 10 out of 10. I also found the Great Learning platform somewhat confusing and not particularly intuitive to navigate. It took me quite some time to become comfortable using it. Regarding the presentations, throughout my professional career—28 years at IBM and now 4 years at Dynatrace—I have always strongly disagreed with presentations that are filled with text while someone is speaking at the same time. I simply do not believe that the human brain is capable of reading and listening effectively at once. Therefore, when watching presentations by Dr. Shelby and Dr. William Roncal, you inevitably have to choose between listening to the speaker or reading the slide. The way I managed this was by pausing the video, reading each slide carefully, and then resuming the video to listen to the explanation. Although this is a tedious process, it greatly improves comprehension. I believe it would be valuable to hear Dr. William Roncal's and Dr. Shelby's thoughts on this aspect of the course. As for the course material and the knowledge shared, I think they are excellent. The program certainly requires significant effort from students, especially for those of us who do not always have a great deal of available time. However, I consider that to be the student's responsibility rather than the course's. Overall, I am extremely satisfied with the knowledge that was shared. My overall assessment of the course is highly positive, and I firmly believe that enrolling in it was the right decision. Regarding the Program Manager, I have nothing but praise. Isha has been incredibly supportive, always attentive to the students' needs, proactive in keeping everyone informed, and committed to resolving any issues that arose. A perfect 10 out of 10 for Isha. One thought that still remains with me, and which I have not yet fully resolved, is whether it still makes sense to spend so much time learning Python at a moment when artificial intelligence can already produce code much faster and often with far greater expertise than most individuals. I believe that, going forward, professionals may benefit more from specializing in harness engineering than from focusing exclusively on learning Python. Finally, I would like to apologize if my honesty seems too direct, but I understand that feedback is requested with the goal of improving the program. I would also like to thank you personally for your professionalism and kindness. My sincere thanks to Hassam for his outstanding attitude, patience, and commitment to teaching—we learned a great deal from him. I would also like to thank the Johns Hopkins University faculty for sharing their tremendous knowledge and expertise with all of us. I would truly love to continue taking more courses because learning is one of my greatest passions. I would even be happy to help promote these programs throughout Latin America. The internet is full of companies offering low-quality courses, whereas this Johns Hopkins University program is, in my opinion, among the very best available on the market and demonstrates an outstanding level of professionalism. If I had to rate the course on a scale from 1 to 10, I would give it a 9.8. The remaining 0.2 simply represents opportunities for improvement. Sincerely, Miguel A. Contreras

Read More

Johns Hopkins University offers a Certificate in Agentics that is a serious and extensive 16-week program requiring, among other things, three significant projects. The program includes pre-recorded lectures by faculty and mandatory weekly in-person meetings in small groups with industry experts. The code is in Python using LangChain and related tools. Johns Hopkins University has a reputation to maintain, and you can see that in the high quality of the offering. This is a serious program, requiring 1 to 2 hours of work per day to get the most out of it. The projects are challenging, and by the end, you have a solid understanding of Agentics, even though in some ways you are just beginning your journey because AI in general, and Agentics in particular, are rapidly evolving. Most importantly, the faculty updates the course regularly, keeps it current, and does a tremendous job of presenting the material. Highly recommended.

Read More

I've learned a lot about Agentic AI through the Agentic AI Certificate Program from Johns Hopkins University. I recommend this course to anyone who wants to keep their technical skills up to date.

Read More

My thoughts on the Agentic AI certification at Johns Hopkins University so far: I'm really happy with the content. I've already been applying everything I'm learning to my daily work, especially since I can now build and deploy agents designed around these core pillars: policy layer, tools and skills, memory and knowledge, guardrails, and observability. The knowledge in LangGraph has been particularly useful for understanding how to steer agent behavior in production at an enterprise scale. Based on what I've learned, I've already managed to build and launch a few agentic solutions into production, from prompt engineering and multi-agent systems all the way to agent evaluation and deployment using LangGraph.

Read More

This certificate has been awesome and fantastic for me. I was motivated to join the JHU Agentic AI program because I wanted to go beyond the general conversation around Agentic AI and Generative AI and develop a deeper, more structured understanding of how autonomous and multi-agent systems can be designed, orchestrated, evaluated, and applied responsibly in real-world contexts. Given my professional focus on digital transformation, data, analytics, and AI-enabled business innovation, I saw this program as an opportunity to strengthen both my technical perspective and my ability to translate AI capabilities into practical business value. My key learnings centered on understanding agentic AI as a system, not just as a model or a prompt. The program helped me better understand agent architectures, multi-agent collaboration, retrieval-augmented generation, tool use, orchestration, evaluation, safety, governance, and the importance of human oversight. One of my most important takeaways was that successful agentic AI solutions require more than autonomy; they require clear objectives, well-designed workflows, reliable evaluation mechanisms, guardrails, and measurable outcomes. The program has helped me significantly in my professional journey by giving me a stronger framework to design and discuss AI solutions with both technical and executive audiences. It has improved my ability to identify high-value use cases, assess feasibility and risk, and think more critically about how agentic systems can be deployed in enterprise environments. It has also reinforced my belief that AI should be applied not only as a productivity tool, but as a strategic capability that can transform operating models, decision-making, customer experience, and innovation. A memorable aspect of the program was the opportunity to connect concepts with hands-on projects and real-world scenarios. Working through practical assignments made the learning experience much more meaningful and helped me internalize how agentic AI systems behave, where they can create value, and where they need careful governance. Overall, the program was intellectually stimulating, highly relevant, and directly applicable to the work I do with organizations exploring the next stage of AI adoption. I am recommending this program to my colleagues, friends, and customers.

Read More

Happy to share—this was genuinely one of the more useful things I've done for myself professionally this year. At my day job at CSL Behring, I lead our automation and AI governance work, and I had already started building an AI assistant for my team on my own. I wanted real training to support what I'd been figuring out by trial and error, and I wanted to actually understand how these AI agents work under the hood. The course built up in a logical way—starting with how to talk to an AI model to get useful answers, then how to connect AI to your own documents and data so it can pull in real information instead of guessing, and finally how to get an AI to plan steps and take actions instead of just answering a question. The two hands-on projects were the best part. In the first, I built a tool that ranks companies as investments by combining real financial numbers with information pulled from their strategy documents. In the second, I built an AI agent that could research a stock on its own—pulling live data, doing the reasoning, and giving a recommendation. Actually building something and fixing it when it broke taught me way more than any lecture could have. It maps directly onto what I'm doing at work. Things I learned in the course—like how much the quality of what you feed the AI matters, and when adding more AI helpers actually makes something better versus just more complicated and expensive—are shaping real decisions I'm making right now on the assistant I'm building for my team. One memorable moment was hunting down a bug in my second project that turned out to be a single letter typed wrong. It was a good reminder that even with all this fancy AI stuff, it still comes down to old-fashioned attention to detail. Really glad I did this.

Read More

I would highly recommend this program to anyone looking to build expertise in Agentic AI, whether they are architects, engineers, data scientists, technology leaders, or AI practitioners. This is not just a theoretical course; it provides a balanced combination of academic depth, practical application, industry relevance, and hands-on learning. The program equips learners with the knowledge and confidence needed to design, evaluate, and deploy enterprise-grade Agentic AI solutions. If you are serious about understanding where AI is heading and want to develop the skills needed to lead the next generation of intelligent systems, this program is an excellent investment in your professional growth. Finally, I would like to thank the entire Johns Hopkins University, Great Learning, and Olympus Learning teams for their commitment, encouragement, support, and inspiration throughout this remarkable learning journey.

Read More

I recently completed Certificate Program in Agentic AI to gain knowledge and understanding of agents and prompt engineering. I believe this course helps me better understand what it means to implement agents in an environment, maintain implementation as the product scales, and stay ethically mindful so that user information and usage of these agents are protected. Having support from the Program Manager was greatly appreciated. All the mentors involved in weekly mentored learning sessions and coursework were very knowledgeable in this complex subject.

Read More