I've spent my career building software and leading engineering teams. From my early days as a software engineer to my current role as a Chief Technology Officer, I've always believed that the best leaders stay connected to the code. When the AI and machine learning wave started turning into a tsunami, I knew I couldn't just lead from the sidelines. I needed to get my hands dirty.
That's what led me to enroll in the 6-month AI & Machine Learning bootcamp through LSU Online and Fullstack Academy. It was a rigorous, deep dive into everything from building models to advanced prompt engineering. The late nights and complex assignments were a humbling reminder of my time as a junior software engineer, but the experience delivered some powerful lessons that apply just as much to the C-suite as they do to the command line.
Here are my biggest takeaways.
Your Model Is Only as Good as Your Data
This was a lesson I learned early in the curriculum that remained true through the final project. I've architected my share of data warehouses and reporting systems throughout my career. I already knew that data was important. But building ML models gives you a deep understanding of why. An elegant algorithm is useless without sufficient, high-quality data. Your primary focus must be on the data pipeline, its cleanliness, and its relevance. Without that foundation, you're just wasting compute cycles. For any leader looking to implement AI, my first question is now: "What is your data strategy?"
Document Your Experiments, Relentlessly
The bootcamp forced us to run countless experiments, primarily through hyperparameter tuning and workflow adjustments. I quickly learned a hard lesson after a few long-running jobs ended in a runtime error or OOM, wiping out hours of progress. Disciplined checkpointing and meticulous documentation aren't just academic exercises. They are survival tactics. In a professional environment, this translates directly to operational excellence. It's the difference between repeatable success and a series of one-off "magic" results that no one can explain or replicate.
Start Small and Iterate
The temptation to build a complex, world-beating model from day one is strong. It's also a massive trap. The most effective approach, much like in agile software development, is to start with the simplest possible solution and iterate. Get a baseline. See how it performs. Then, layer in complexity incrementally. This iterative loop saves time, clarifies thinking, and prevents the kind of premature optimization that can kill a project before it even gets off the ground.
Not Every Nail Needs an AI Hammer
One of the most valuable lessons was learning when not to use AI. As a technology leader, it's my job to align our technical strategy with real business challenges. The hype around AI can create pressure to apply it everywhere. But a simple script or a well-designed traditional system is often a faster, cheaper, and more reliable solution. The real skill is identifying which problems are genuinely suited for a machine learning approach. The goal is to solve business problems, not just to keep up with trends.
Use LLMs as a Learning Partner
I used commercial LLMs throughout the course as a partner. When a concept was confusing, I'd ask for a simplified explanation or a practical code example. This on-demand, personalized instruction accelerated my understanding in a way that traditional resources couldn't. It's a powerful demonstration of how to use AI to master AI.
Going back to a formal learning environment after years in a leadership role was an investment, but it paid off. It sharpened my technical skills, reinforced core engineering principles, and gave me the hands-on perspective I need to lead our AI initiatives effectively.
