Post by Codecademy

We challenged learners in our Machine Learning Foundations for Beginners bootcamp to build a neural network from scratch that could classify handwritten letters (A-Z), then test it on completely unseen, hand-crafted data. Linda Walton told her instructor she definitely wouldn't win. She came in first. Linda’s model achieved 92.3% accuracy on real-world handwriting data, using just 274K parameters out of a 1M budget. Her instructor called it the most efficient submission in the batch, crediting smart architectural choices (dual convolution blocks) and diverse data augmentation for keeping her gap between test and real-world performance to just 2 points. The most impressive part? Linda did all of this while working full-time as an ML Engineer. "It hasn't been easy," she says. "Transitioning into this branch of engineering while balancing my work has come with plenty of impostor syndrome moments." Linda kept going anyway. In 4 months, she passed the IBM Machine Learning Professional Certificate, started prepping for the AWS ML Associate exam, and earned a spot in Johns Hopkins' MS in Artificial Intelligence program. Congrats to our winner and every learner building real projects that matter. 🎉 Ready to build your own real-world ML project? See what's coming up next: https://bit.ly/4spuCfa