Deep Learning
My Deep Learning Journey: Lessons from a Master’s in AI at the University of East London
When I first started my master’s in AI at the University of East London, I felt both excited and completely overwhelmed. Deep learning seemed like a magical black box, full of terms I barely understood—neural networks, backpropagation, transformers—it was all a bit intimidating.
But as I navigated coursework, hands-on projects, and late-night coding sessions, I realized that deep learning is less about memorizing formulas and more about curiosity, persistence, and trial-and-error.
The Struggle Was Real
I remember my first attempt at implementing a neural network in Python. What should have been a straightforward task quickly turned into a debugging nightmare—NaNs popping up everywhere, gradients going rogue, and models refusing to learn. At first, I felt frustrated, but over time I came to appreciate those moments. Every error was a lesson, every failure a stepping stone.
One thing became clear early on: learning happens when you try, fail, and try again.
Hands-On Projects That Changed Everything
The turning point for me was projects. From training a simple feedforward network on MNIST to experimenting with convolutional neural networks for image classification, I gradually gained confidence.
Working on NLP tasks was a revelation. Tokenization, embeddings, and transformer models weren’t just buzzwords anymore—they were tools I could experiment with. Training even a small transformer model helped me understand the power and complexity of modern AI.
I realized that starting small, understanding each step, and then scaling up was far more effective than trying to tackle huge models from the outset.
Lessons Beyond the Code
Deep learning isn’t just about writing models—it’s about thinking like a researcher. At UEL, we were encouraged to read papers critically, replicate experiments, and contribute to open-source projects. I learned to question assumptions, analyze results carefully, and value reproducibility—skills that will stay with me long after graduation.
Another important lesson? Community matters. Discussing problems with classmates, participating in hackathons, and seeking guidance from mentors accelerated my growth more than any solo study session ever could.
Advice for Aspiring Deep Learning Practitioners
Start with projects, not just theory. Coding simple models teaches more than memorizing formulas.
Embrace failure. Every underperforming model is a chance to understand the subtleties of AI.
Read papers critically. Understanding why an architecture works is more valuable than just knowing that it does.
Engage with the community. Online forums, GitHub projects, and study groups can accelerate learning.
Stay curious. AI is evolving fast—keep exploring and experimenting.
Looking Ahead
Deep learning has become more than just a field of study for me—it’s a lens through which I see problems, from AI research to real-world applications in healthcare and robotics. My master’s at the University of East London gave me the foundation, but the journey is just beginning. Every day brings a new opportunity to learn, experiment, and push the boundaries of what machines—and we—can do.
