From Aerospace Dreams to AI Domination Behind the Rise of Labelbox

Machine learning projects rely heavily on data. While algorithms and computing power garner much attention, the foundational reality is that artificial intelligence requires vast amounts of meticulously labeled data to function effectively. This truth was particularly evident in the early 2010s when Manu Sharma, a young aerospace engineer at Embry Riddle Aeronautical University, was experimenting with neural networks.

"At that time, working with neural networks was still archaic," Sharma recalls. "This wasn't even too long ago, around 2009 and 2010, but one of the best ways to work with neural networks was to use MATLAB or Simulink software packages that are widely available at educational institutions."

What intrigued Sharma was the way these early AI systems learned. "I was very intrigued by these systems because you didn't have to learn equations to figure them out or train them. You just fed them data and it would start mimicking behaviors it would see from patterns in the data."

During his university years, Sharma spent a significant amount of time testing and developing neural network solutions for the aerospace industry, with a focus on autonomous flight. His classmate, Brian Rieger, shared similar interests, and the two collaborated on multiple projects before their professional paths diverged—Sharma taking product roles in Bay Area startups and Rieger joining Boeing as a data scientist in Texas.