Hey everyone,
I wanted to share a quick reflection on how much the field of AI research has changed recently and see how others are navigating it.
Looking at the landscape today, there has been a massive shift since LLMs and VLMs took over the market. It used to be that the work was centered around the foundational process of data cleaning, model building, hyperparameter tuning, performance evaluation, and deployment. I feel that the focus has lately moved away from traditional model development toward application building and deployment.
Because I have spent most of my time focusing on model architecture, evaluation, and performance analysis, I have found it challenging to keep up with this transition. Trying to establish my career right at the moment this industry-wide pivot is happening has been a steep learning curve.
I would love to hear your thoughts on how you are navigating this shift and what has helped you adapt.
Hi Jasmeet,
You have mentioned a valid input towards whoever that is working towards cracking an AI Engineering role. My approach is pragmatic here. If I directly jump into application side of LLMs then I’d have a lot of knowledge gaps which would stall me in efficient problem solving. My learning is a mix of fundamentals and the current agentic engineering and orchestration. I’m open to feedbacks for this comment of mine if I do have limited exposure.
Hey @noishey.
I totally agree with your point. Jumping straight into application building can be limiting for someone wanting to establish career in AI research.
I follow a very similar hybrid approach, balancing practical application building with foundational principles. It kind of helps you understand why things work, not just how to plug them together. Your strategy sounds incredibly solid!
Also, welcome to the community! 
I believe it also depends on the direction you want your career to take and how deep you want to go in the field. However, I agree that having a solid understanding of the fundamentals is never a bad thing—it actually becomes an advantage.
A good analogy is a mechanic. Even if modern cars have sophisticated software and diagnostic tools, a mechanic still needs to understand how the engine and its components work. Likewise, understanding the foundations of machine learning helps you troubleshoot, evaluate, and build better AI systems, even if much of today’s work focuses on LLM applications and AI engineering.
That’s a great analogy @gent.spah, I completely agree with your point.
What I find a little challenging is finding the right balance between application focused work and the research. I think with job requirements growing so much, it feels important to be comfortable with both building systems and understanding how they work.
Do you have any project suggestions that combine both?
I don’t have a specific project in mind that naturally combines both research and application, except for research-oriented projects themselves. In most research projects, you’ll still build on existing frameworks, libraries, or prior work while contributing something new, so they tend to sit at the intersection of both worlds.
I also agree that job expectations keep growing. At the same time, there’s a limit to how much any one person can learn. As the Zen saying goes, you have to empty some water from the glass before you can pour in more. Competition and ambition never really end.
My approach is to focus on strong fundamentals and stay aware of what’s out there. You don’t have to master every new tool immediately. When a project or job requires a specific technology, you can learn it then. Treat learning as a long-term process rather than a race—it becomes much more sustainable that way.
Hello @Jasmeet_Singh2,
I recommend that you go over the DLAI short courses catalog and see, at least, which titles are interesting. Most, if not all, of these courses are application-focused: you are going to achieve something in the provided Jupyter Notebook interface.
The catalog also shows the names of the associated organizations, so you can easily get an idea of the tech stack or the kind of application to be built. Some of these courses come with explanations for what is under the hood, but if the level is not deep enough for you, the good news is that the courses are generally short (thus the name), so you would have plenty of time to ask an AI chatbot for more references while still being able to finish them in time.
The exciting part of this learning platform is that you can expose yourself to a wide variety of topics covered by industry experts, before you decide where to dig deeper. This way, you may find a better balance that is more suitable for you.
Cheers,
Raymond