FastAPI & Machine Learning: Build a Banking Fraud Detection (Udemy.com)

Its really a very good tutorial.

Cons 1. Lack of Depth: The course focuses on "pure code typing" rather than teaching engineering. The instructor narrates exact his typing every single line, but rarely explains the underlying "why" or "how." at both high and granular level. He seems knowledgeable but lacks teaching skill. It feels like a copy-pasted production project with shallow explanations. A major example of this oversight is the missing evaluate_fraud_model_performance() Celery task in ./backend/app/core/task/ml.py, which is absent from both the course and the GitHub repository. 2. Weak AI/ML Sections: (You will curse and give up this entire section if you don't have extensive prior AI/ML knowledge and experience.) The instructor simply reads from two summary markdown files as if that will help us any true understanding. I spent double the course time doing outside research just to understand what over 3,000 lines of code in this section are doing. 3. For those on Windows + Docker Desktop has to run the containers on WSL as volume mount does NOT work on standard windows file system. In addition, you will have to toggle configurations between WSL and localhost so that Traefik can auto detect other containerized services. Good luck on this lol. 4. Distracting Production: The constant screen zooming, highlighting, and magnification—combined with low-volume mumbling and distracting background music—is incredibly taxing. It causes significant visual and auditory strain, making it difficult to maintain focus. 5. Ineffective Delivery: The instructor prioritizes visual effects over clear communication. (Please study top-tier instructors like Maximilian Schwarzmüller, Ed Donner, or Stephen Grider to see how complex technical content should be delivered.) 6. Repetitive Verbal Tics: I lost track of how many hundreds of times he said "it’s going to be," "like that," and "just like that" as he types every single line. These verbal crutches are extremely bothersome and make the learning experience mentally exhausting. Pros 1. Breadth of Concept: Be prepared to spend at least 100 additional hours self-studying to truly grasp the implementation. This course is a solid choice if you want exposure to full modern fintech stack. 2. Unmatched (but Unrealistic) Scope: The sheer volume of knowledge condensed into this single course is mind-boggling. It covers several distinct specializations: -Distributed Systems: Celery, RabbitMQ, Redis, worker management, rate limiting, and beat scheduling. This alone should be 30 hour course! -Backend Engineering: FastAPI, SQLModel, async PostgreSQL, and complex financial transaction flows. This alone should be 30 hour course! -ML Engineering: Feature engineering, gradient boosting, pipelines, and data leakage prevention. This alone should be 30 hour course! -MLOps: MLflow, model registries, drift detection, and retraining pipelines—a cutting-edge field in itself. A relatively new field that barely existed 5 years ago! -Production Deployment: Yet another full specialization. This alone should be 10 hour course!

This course is amazing.

Alpha has done a great job in covering the entire course to develop banking APIs using Fast API while making use of multiple tools. My only concern is that the course expects one to watch through the entire code being manually written which is time taking. Instead, the repo can be provided by module as the code changes in the same file as the course progresses. The repo which has been shared works as an end exercise which will get everything working except the section 19. So I think Alpha should create the course in such a way that one should be quickly be able to make progress. The course takes 33 hours because one has to listen to each line of code. Some may prefer to first set up the environment, and then walk through the code and do the testing. This may take less time to grasp the same. But with his code approach, you get to see Alpha troubleshoot, which is also a positive side of his way of teaching. I see a mention of nothing is working, but everything worked for me, especially if you maintain the same version of libraries as in the requirements.txt

l'introduction du cours est plutôt prometteur j'espère que la suite sera autant

I took a similar course from Alpha but using Django instead of FastAPI. This is an improvement of the same use case, but using async programming and machine learning. Both courses are very good, well structured, with a lot of useful information, but not for beginners. In my case I have learned a lot, and it has been very useful to detect the parts of FastAPI and ML knowledge that I must improve. Summarizing, a very good course, but you should have some background in FastAPI and ML before doing it. Congratulations to Alpha for a well done job.

So far the course is been awesome. There are some stuffs that would be changed, like use a TDD approach and use just one package manager instead of use Pipenv and requirements, when could use just the pipenv, poetry or uv and using the groups to install on each, but those are just suggestions, it does not affect the course content.

This course offers an impressive depth of content and delivers substantial value, especially now that it has been fully completed. Some of the earlier 1-star reviews reflect a time when the machine learning modules had not yet been released. Today, the course is complete and exceptionally well-structured. It is important to note that this course is not intended for beginners. It assumes a working knowledge of FastAPI and a foundational understanding of machine learning concepts. For example, prior experience with tools like MLflow, along with familiarity with classification and regression problems, is essential to fully understand the reasoning behind certain design choices, such as the use of Gradient Boosting or specific deployment workflows. If you already have a basic background in FastAPI and some experience in machine learning, you will likely find this course both enjoyable and highly rewarding. Alpha guides you through building a fully functional backend application from scratch using a thoughtful and scalable structure. While this course will not prepare you to launch a fintech business, particularly given the regulatory complexities in Europe, it will absolutely enable you to build and deploy your own machine learning-powered applications. This includes training, evaluating, and serving your own models with MLflow, which goes far beyond simply integrating large language models via API calls. One final note: this course focuses solely on the backend (API). If you are looking for a full-stack solution, Alpha offers another course that may be a better fit.

There is so much to learn and I hope it meets the expectation.