Why AI Needs Great Software Engineers

Modern software engineers need to be familiar with generative AI and predictive analysis. Whether you already have a data science background and want to become a machine learning (ML) engineer, or are just starting in tech and want to dive into an exciting, cutting-edge career path, learning from a coding bootcamp will ensure you’re prepared for the workforce in 2024. Will Sentance, CEO of Codesmith, shares how they’ve integrated AI into their updated Full Stack Bootcamp curriculum, and how software engineers are embracing AI in 2024. 

Let’s address the doomsday headlines: Will AI replace software engineers? 

There is an overused phrase: “AI is eating the world.” In reality, artificial intelligence is opening up new realms of opportunity by enabling software to deal with things that it previously couldn’t. Software will now be able to solve problems that involve uncertainty or nuance, like legal advice or healthcare recommendations, and the people building the solutions will be software engineers. 

Even a technical product manager won’t have sufficient expertise — it’s got to be someone who understands what’s going on under the hood. 

Thanks to AI, many more of our societal problems will be solved through the software layer. This why it’s so important for the people becoming technologists and software engineers to have lived experience as technology users. The people building the tech need to represent society at large, particularly when you’re expanding the realm of things that can be solved with software. 

For example: We have a grad who was a nurse for nine years and who now works at a medical staffing agency as a software engineer. This is someone who has lived it and is now leading it. Schools like Codesmith offer a path for people — whatever stage they’re at in their lives — to enter into a position of autonomy over tech, rather than staying on its receiving end.

Are coding bootcamps irrelevant now that AI has become increasingly ubiquitous?

Coding bootcamps have to evolve, there’s no doubt about that — which is why Codesmith has intentionally incorporated AI into our curriculum. But the core of everything is still software engineering. If you’re a data scientist, for example, your job is to build a model that can be used to make predictions, but what turns this model into something that businesses and individuals can use depends on software engineering. 

While a coding bootcamp won’t make you a machine learning engineer (unless you arrive with a heavy data science background), what you can do in a coding bootcamp is become a software engineer who’s ready and willing to work closely with ML engineers to integrate new features of AI, which involves a mixture of data science, ML engineering, and software engineering.

If you’re going to learn AI, it has to be built out end-to-end as part of your growth as an engineer, not just as a trivial add-on. At Codesmith, we are lucky to have gifted curriculum developers, and we’ve enjoyed augmenting the curriculum to reflect the needs of the modern full stack developer. The updated Codesmith program integrates front end, back end, infrastructure, dev tools, AI software engineering support tools, and ML-adjacent engineering.

But there’s much more that a coding school can do for individuals and for society at large. Getting back to my earlier point about representation, we see how the algorithms of many tech platforms (such as Facebook and TikTok) are of global political importance. And yet these algorithms are essentially black boxes. You need the people running these platforms to have been on the receiving end of technology, or else you’ll end up with that same thing that happened in the mid-2010s with social media.

What I love about coding bootcamps is that they nurture people who didn’t wield technology from Day One. You need some people running this stuff who have had technology wielded on them. Given how hard it is to change direction past 18 years old, you bet these schools will become even more important when the things being built matter as much as healthcare, financial recommendations, and global policy.

How are software engineers embracing AI in 2024? Which AI tools are they leaning on or has it been more of a mindset shift?

There are a few main ways software engineers are embracing AI this year. Some people are working as straight-up machine learning engineers. Others are integrating APIs (like those from OpenAI or Anthropic) or libraries like Langchain into the software architecture of a product they’re helping to develop. Still others are able to reason through ever more complex tasks with the help of AI tools like ChatGPT and Github Copilot. This lets them apply their understanding of how the AI is working under the hood to make predictions based on input. 

Regarding whether or not it’s a mindset shift, the core things that make someone a great software engineer — like problem solving ability and technical communication — won’t change. The engineering approach may change slightly, given that the mental models — the principles that give you an understanding of how the code runs — are expanding. With AI, these new mental models are going to involve concepts like prediction and uncertainty. 

Things change, but software engineering is still the incremental grind of building out products and features without breaking the existing functionality. It’s not going anywhere!

What is a high-level definition of predictive analytics? 

A data scientist’s job is to build a model, which is a collection of weights (composed of the code and the steps taken with any previous data) that allow them to make predictions about new situations. This is essentially predictive analytics. For example, you might have a model that looks at previous users' purchase history or searches. According to that, you’ll make predictions about what the next user might search for or purchase. 

With predictive analytics, you need to understand the math behind building models to make good predictions. With generative AI, you need to understand the math and data science behind it to make sure it's fast and inexpensive.

The data scientist does this in a very small scale way, which is what makes it so powerful. They’re able to turn this data into something that users, businesses, and others can use. It’s important to note that the software engineer is going to be part of building that; they will be helping with model deployment, which looks a lot like server development or back end development. 

Are employers today expecting incoming tech hires to know how to use AI methods and tools?

Are they expecting it? Hard to say. Is it an edge? Sure. Obviously, the ML engineer has to understand the model, the math, the analytics, and the software engineering. I’m also seeing a lot more of the “fusion” roles, in which software engineers are either customizing or fine-tuning ChatGPT-based prediction tools to generate better predictions. They’re taking an LLM (large language model) tool and building a product around it. I see up to 10% of people going into those roles. 

Employers are hiring for a range of roles that involve at least some interaction with AI. These roles fall into three main buckets:

• Machine Learning (ML) Engineer — ML Engineers have to understand the model, math, analytics, and software engineering. We anticipate seeing a lot more bootcamp grads getting hired for these roles, especially those who come in with a data science background and want to fill in gaps by learning software engineering at a coding bootcamp.
• GenAI Engineer — This is the intersection of using GPT-based prediction tools like LLMs and either customizing them with your software engineering knowledge or fine-tuning them to generate better predictions in order to create quality products. 
• Human Feedback Consultant — The Human Feedback Consultant ensures that the code being generated by models follows best practices and can solve problems. In these roles, you’re working with an AI model, making predictions, and assessing its output based on your expert knowledge as an engineer.

Will students at Codesmith learn artificial intelligence methods and tools? 

Codesmith recognizes the full journey of understanding, taking learners through the total pipeline, rather than segmented add-ons that cover artificial intelligence. At Codesmith, we cover:

• The mental model of domain (programming and how it runs; LLMs and how they work)
• Implementation details (the exact AI tools to use, such as LangChain)
• React in front-end development
• Node in back-end development

Codesmith has pre-program workshops that dive into the more challenging parts of the program. We believe that learning includes principles and mental models and not just tools, which will change frequently. 

What is your advice to recent bootcamp grads on the job hunt now? What can they do to differentiate themselves in a competitive tech talent pool?

First, understand the principles of software engineering and machine learning, then move into tool implementation. Build projects with them, and build tools for other developers to use the mainstream tools. For instance, you could build a visualizer to let other developers better understand how their React code base is working. Make sure you learn how to talk about these topics and the choices you make as an engineer.

Find out more and read Codesmith reviews on Course Report. This article was produced by the Course Report team in partnership with Codesmith.