It started with building things
One of the things I enjoy most about programming is how quickly an idea can become something real.
You can start with a problem, a vague idea of how to solve it, and a computer. A few hours later, you might have something that actually works.
That’s what initially drew me toward data science and software engineering.
At UC San Diego, I’ve spent a lot of time learning about statistics, machine learning, and programming. But some of my favorite experiences have happened outside the classroom, when I’ve had the freedom to build something simply because I thought it would be interesting.
Hackathons were a big part of that.
At UCSD DataHacks, my team worked on a wildfire simulation application. The idea was to model how a wildfire could spread across a geographic area and visualize the direction and progression of the fire.
It was an interesting problem because it brought together data, geography, simulation, and software development.
Our project won the cloud development category, which was exciting. But what stuck with me wasn’t just the result. It was the experience of turning an abstract problem into something people could actually interact with.
At another hackathon involving LPL Financial, my team built an automated form-filling application. Instead of manually entering the same information into different forms, a user could provide their information and have the system populate the relevant fields.
These were different projects, but they reinforced the same idea: software is interesting because it can make complicated things useful.
AI made building easier
Around the same time, AI coding tools were becoming increasingly capable.
Suddenly, you didn’t have to spend nearly as much time figuring out every implementation detail before seeing an idea take shape.
You could describe what you wanted, generate an initial implementation, test it, and keep refining it.
For someone who enjoys experimenting with ideas, that was incredibly appealing.
I’ve used AI-assisted development while working on personal projects, including web applications and tools that I wanted to make more polished and useful.
It made the process of building feel much more accessible.
But it also introduced a question that I kept coming back to.
If an AI can generate code that works, how much of that code do I actually understand?
Making something work isn’t the same as understanding it
There’s a difference between getting a program to produce the expected result and understanding why it produces that result.
That distinction becomes especially important in data science.
A machine learning model might generate predictions. A visualization might look convincing. A web application might behave exactly as expected.
But those outcomes don’t automatically tell you whether the underlying implementation is correct.
You still need to understand the assumptions.
You need to know where the data came from, how it was processed, what the model learned, and whether the evaluation actually measures what you think it measures.
You also need to understand what happens when something goes wrong.
I’ve become increasingly interested in that part of engineering.
Not just building the visible result, but understanding the systems that make the result possible.
The interesting problems are underneath the surface
When I first became interested in artificial intelligence, I was naturally drawn to what models could do.
The ability to generate text, write code, recognize patterns, and solve complicated problems is fascinating.
It still is.
But the more I work with software, the more interested I become in the infrastructure underneath those capabilities.
How do you make a system reliable?
How do you design software that remains maintainable as it grows?
How do you know whether a model is actually performing well?
How do you move from an interesting prototype to something that people can consistently depend on?
These questions are less flashy than watching an AI generate an entire application in a few minutes.
But I think they’re just as important. Maybe more important.
What I’m interested in now
I still enjoy building with AI.
I don’t think there’s anything inherently wrong with using tools that make development faster. If anything, I think they’re going to become a normal part of how software gets built.
But I don’t want to become dependent on tools I don’t understand.
I want to become better at the fundamentals: programming, statistics, machine learning, software architecture, and the systems that connect everything together.
I’m particularly interested in AI research, inference, and software infrastructure.
I like the idea of working on systems where performance, reliability, and technical decisions actually matter.
And I want to keep building projects that force me to learn things I don’t already know.
Still figuring it out
I’m still a student.
There are plenty of things about software engineering and machine learning that I don’t understand yet.
But I think that’s part of what makes the field interesting.
Every project exposes another problem to solve, another concept to learn, or another assumption worth questioning.
AI has made it easier than ever to build something.
For me, it’s also made understanding what I’m building feel more important than ever.
And that’s something I want to keep getting better at.