Learning to Work with AI: Reflections on Software's Next Chapter

Jeff Larkin’s Visit & Reflecting

This past Tuesday, our class had the opportunity to hear from Jeff Larkin, a director at NVIDIA with deep experience in AI and software engineering. His talk was not just about cutting edge technology. It was about how AI is actually being used today, how engineers think about it, and what it means for us as students preparing to enter the field.

What stood out most to me was how practical and grounded his perspective was. AI was not presented as magic or hype. It was presented as a tool that is powerful, evolving, and deeply integrated into real engineering workflows.

Alongside Jeff Larkin's talk, I also listened to John Miller on the Gettin' to Market podcast, where he discusses Personal Brand with Mary Scott Van Arsdale. The assignments asked me to synthesize these perspectives and reflect on what they mean for the future of software engineering.

What I Took Away From Jeff Larkin’s Visit

Defining AI and Learning

We started by breaking down what AI and learning actually mean. At its core, learning in neural networks comes down to backpropagation. If a model predicts z but the correct answer is y, the system adjusts the neurons so that next time it is more likely to predict y. That simple feedback loop of predict, compare, and adjust is what allows models to improve over time.

It reminded me that beneath all the complexity, AI systems are built on structured mathematical feedback and optimization.

AI for Weather Prediction and Earth-2

One of the most surprising things I learned was how AI is being used for weather prediction. Jeff talked about NVIDIA Earth-2, which is essentially a digital twin of Earth.

A digital twin is a virtual replica of a real world system that can simulate and predict behavior. In this case, Earth-2 uses AI and computer graphics to simulate climate and weather systems at an unprecedented scale.

Traditionally, weather forecasting relies heavily on statistical and physics based models. These methods are powerful but computationally expensive. Earth-2’s AI driven approach is not only more computationally efficient and uses less energy, but it is also more accurate in certain cases. Jeff showed an example involving a storm near Taiwan where the AI model was able to account for historical outliers that traditional models typically struggle with.

This demonstrated something important. AI is not just faster. It can sometimes see patterns that conventional methods miss.

AI and Computational Fluid Dynamics

Another powerful example involved computational fluid dynamics, often referred to as CFD. Traditional CFD simulations used heavily in aerospace and automotive engineering can take 8 to 9 hours to run even on extremely powerful computers.

With AI based simulation models, engineers can create emulations of these CFD simulations. These AI models do not replace the full CFD process, but they dramatically accelerate early stage testing. Engineers can iterate through design changes much faster, using AI as a starting point before running full detailed simulations.

This changes the design workflow. Instead of waiting hours between iterations, engineers can experiment rapidly and refine ideas before committing to expensive computation. It is not about replacing physics based modeling. It is about augmenting it.

Digital Twins

The concept of digital twins kept coming up throughout the talk, and this was actually the first time I had heard of the concept. A digital twin is more than just a simulation. It is a dynamic, data informed replica of a real world system. Whether it is Earth’s climate, an airplane wing, or an industrial process, digital twins allow engineers to test scenarios, predict outcomes, and optimize performance without physically altering the real world object.

Learning about this for the first time made me realize how much engineering is shifting toward virtual experimentation before physical implementation. This feels like a major shift in how engineering problems will be approached over the next few years.

The Future of Jobs and AI

One quote that really stuck with me came from Jensen Huang:

That statement reframes the fear around AI. The technology itself is not the threat. Stagnation is. Engineers who learn how to integrate AI into their workflows will have a massive advantage.

Final Reflection

Jeff Larkin’s visit made AI feel less abstract and more tangible. It is not just about chatbots or generating text. It is about accelerating engineering, improving simulations, reducing energy usage, and enabling entirely new workflows.

Over the next two years in software engineering, I think the biggest shift will not be whether AI exists. It will be how effectively we learn to use it. The engineers who treat AI as a collaborative tool rather than a replacement will likely be the ones who thrive.

If anything, this visit made one thing clear. Understanding AI is not optional anymore. It is becoming foundational.

Response to John Miller's "Gettin' to Market" Podcast

Episode: 35 - Personal Brand with Mary Scott Van Arsdale

As a college student, I had already understood the importance of LinkedIn through the networks I have built and the opportunities that came from them. I have been able to find connections to recruiters simply by navigating my LinkedIn network and identifying shared connections. That alone showed me how powerful the platform can be. However, this episode expanded my understanding in ways I did not expect. One of my biggest takeaways was realizing that you do not always have to post constantly to maintain a presence. The discussion around commenting really stood out to me. Contributing thoughtfully to other people’s posts allows you to stay visible, add value, and build credibility without the pressure of always creating original content. I also gained a new perspective on maintaining a personal brand even after securing a job. Previously, I thought personal branding was mostly relevant during a job search, but this episode made it clear that it should be consistent and ongoing. I have already seen the impact of consistency in my own experience. As a teaching assistant with significant experience in that area, I have had recruiters reach out to me about teaching positions because my profile consistently reflects that focus. This episode helped me connect the dots between clarity, consistency, and opportunity, and it reinforced that personal branding is not about self promotion but about being intentional with how you present your strengths over time.

Two Diary Entries: AI-Powered Future (50 Years from Now)

Diary Entry 1: A Negative Future

Dear Diary,

I can't believe this is my 250th interview and I still haven't landed a job. The majority of SWE positions are now filled by AI with just one person overseeing the work. Teams that once needed 50 engineers now only need one. I don't know how many more rejections I can take. The job market is brutal right now.

I've been seriously considering switching careers entirely. Maybe becoming a cook or a ski instructor would be better. At least I'd be able to find and keep one of those jobs. I'm regretting the six years I spent in college studying computer science. What was the point of all that? If I had studied something I enjoyed like art, or gone to culinary school, I could have been a head chef at some fancy restaurant by now. This future is bleak. I'll keep you updated on what happens next.

-your sad SWE

Diary Entry 2: A Positive Future

Dear Diary,

I can't believe how much I love my job now. The parts I used to dread, specifically those tedious syntax errors, are completely handled by AI. Back in college, I spent countless hours debugging what I thought were complex logic issues, only to discover they were simple syntax mistakes or subtle language quirks I didn't fully understand. Those frustrating days feel like ancient history now. I can focus on the part of computer science I love, designing and solving problems.

Another thing I've come to love is how quickly I can experiment with code and ideas. If I have an idea but I'm not sure how to execute it, I can pair program with AI and prototype it rapidly. It doesn't even have to lead anywhere—the process of trying something out has just become effortless.

-yours SWE

Weekly Progress on Nogramming Assignment

I have decided to switch my idea to creating a podcast where I interview people with different AI experiences to understand their perspectives on the technology. This week, I'll compile a list of potential guests and reach out to gauge their interest.

~Shree