Deep Learning

This course was designed as an intensive, hands-on introduction to deep learning for students who wanted more than a high-level overview. The emphasis was on building working intuition, implementing models from scratch, and understanding how modern deep learning systems are trained in practice.

I co-instructed the course as a subject matter expert and delivered 20 hours of in-person instruction across a one-month, 30-hour program. The cohort included around 40 students, many of whom were transitioning from general machine learning concepts into deeper work on neural architectures.

The course covered:

• neural networks and the fundamentals of representation learning
• backpropagation, optimization, and training dynamics
• convolutional neural networks for image tasks
• recurrent models, LSTMs, and sequence modeling
• generative models including GANs
• explainable AI and practical model interpretation

My focus was on making the material implementation-oriented. Instead of presenting architectures only as abstract diagrams, I tried to connect each topic to the practical questions students face when they actually train models:

• why a model is not converging
• how to interpret training curves
• when a deeper model helps versus hurts
• how to select hyperparameters under limited compute
• what makes an experiment reproducible and trustworthy

In addition to lectures, I helped design assignments and guided students through project work so that they could apply the material on real examples instead of stopping at theory. The course was especially valuable as a bridge between classroom ML and project-based deep learning.

Teaching this course reinforced my interest in technical mentoring and in making advanced ML concepts accessible without oversimplifying them. It also gave me a deeper appreciation for the difference between understanding a model on paper and learning how to make it work in practice.