𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝗦𝘁𝗮𝗿𝘁𝗲𝗱 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴

Recently, I got a recommendation for the book AI Engineering: Building Applications with Foundation Models by Chip Huyen, and I decided to give it a try.

I finished Chapter 1, and I really loved it. The examples, use cases, and overall flow make the concepts easy to understand.

To complement that, I also started preparing for the AWS AI Practitioner certification. My goal is to not only understand the concepts but also see how they apply in real world systems. I did not want to just read about AI, I wanted to connect it to how things are actually built and used.

𝗔 𝗠𝗼𝗿𝗲 𝗣𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗩𝗶𝗲𝘄 𝗼𝗳 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴

One idea that stood out to me is how AI engineering is evolving in practice.

While traditional machine learning focuses heavily on model development and training, AI engineering today is increasingly centered around how we use and adapt existing foundation models in real world applications.

This shift moves the focus from building models to designing systems. It is less about training from scratch and more about how effectively we can guide models, provide the right context, and evaluate their outputs.

What I found particularly valuable is how this perspective aligns closely with software engineering principles. The emphasis is not just on the model itself, but on how it fits into a larger system and delivers value in a practical setting.

𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 𝗵𝗼𝘄 𝗺𝗼𝗱𝗲𝗹𝘀 𝗽𝗿𝗼𝗰𝗲𝘀𝘀 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲

One concept that stood out to me was how models process text.

Instead of reading full words the way we do, they break text into smaller pieces called tokens. For example, a word like “cooking” can be split into “cook” and “ing.” Even a completely new or made up word can still be interpreted because the model looks at its smaller components.

This approach allows models to capture meaning while remaining efficient, and it helps them handle words they have not explicitly seen before

It also explains why interactions with these models feel flexible and natural.

𝗛𝗼𝘄 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲𝘀 𝗮𝗿𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱

Another important idea is how these models generate responses.

They do not reason in the same structured way humans do. Instead, they predict what comes next based on the sequence of tokens they have already seen. Each response is built incrementally, one step at a time, based on learned patterns.

When viewed this way, it becomes clearer why interactions feel conversational. The model is continuously predicting the next most likely piece of text in a sequence

𝗙𝗿𝗼𝗺 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹𝘀 𝘁𝗼 𝗮𝗱𝗮𝗽𝘁𝗶𝗻𝗴 𝘁𝗵𝗲𝗺

A key takeaway for me is that building AI applications today does not always require training models from scratch.

Instead, much of the focus is on adapting existing models. This can be done by providing better instructions, supplying additional context through external data, or refining the model for specific use cases.

This shift lowers the barrier to entry and enables faster experimentation. It allows developers to focus more on solving problems and less on the complexity of model training

It feels great to understand the nuances behind what we use every day. I use LLMs on a daily basis, and getting to learn what is happening underneath has been fascinating so far.

From a software engineering perspective, this shift is particularly meaningful. It reinforces the importance of building, iterating, and integrating systems rather than waiting for complete theoretical understanding.

It also highlights that many core engineering skills remain relevant. Designing systems, thinking about user experience, and building reliable applications continue to be central, even as the tools evolve.

At this stage, I am continuing the book while also preparing for the AWS AI Practitioner certification. I want to build a solid foundation while staying grounded in practical application.

In parallel, I am exploring ways to apply these concepts in smaller projects. Even simple implementations can provide valuable insight into how these systems behave and how they can be improved.

This combination of learning and building feels like the most effective way to deepen my understanding.

AI is no longer just about models, it is about systems.

Understanding how to use, guide, and integrate those systems is what makes the difference.