Architectures, not just APIs
Implement CNNs, transformers, and generative models from the ground up so you understand what's happening beneath the framework calls.
OCTOBER 2–16, 2026
Deep Learning,
Masterclass
3-WEEK PROGRAM FOR AI/ML PRACTITIONERS
From neural network internals to transformers and generative models. Seven sessions across three weeks, for practitioners ready to master the architectures behind modern AI.
7hands-on sessions
3immersive weeks
01capstone project
About the workshop
Go beyond the
framework call
Deep Learning Masterclass is a seven-session program for people who already write Python and understand core machine learning — and are ready to go deep into the architectures behind modern AI.
Work through neural network internals, convolutional and generative models, sequence models and attention, and transfer learning. Then bring it all together in a capstone project of your own.
Train on real models and data
Work with real vision and sequence datasets and pretrained models, at a pace built for people already writing production code.
Ship a capstone project
Take a deep learning idea from architecture choice to a trained, evaluated model you can showcase.
Speakers
Meet Our Speakers
Learn directly from the researchers and educators leading AI and deep learning at CITY, LUMS.
DK
SPEAKER 01
Dr. Zubair Khalid
Associate Professor, LUMS — Director, CITY
Dr. Zubair Khalid is an Associate Professor at LUMS and Director of CITY, where he advances AI-driven, data- and policy-informed solutions for digital transformation, climate resilience, and urban systems. His broader research spans artificial intelligence, machine learning, systems engineering, computer vision, signal processing, and optimization.
He completed his PhD in Engineering at the Australian National University, Canberra, in 2013, and held positions as Professor at KAUST and AI Resident at Maincode. At LUMS, he teaches Machine Learning courses (AI501/EE 514/CS 535) in the MS Artificial Intelligence program, covering the theoretical foundations and practical applications of supervised and unsupervised learning. His work has been recognized with the ACM Gordon Bell Prize (2024) and contributions to AI for Industry R&D.
FOCUS AREASAI strategy & use-case prioritization · AI on edge devices & on-prem AI · Digital twin & modeling · Governance: privacy, IP & accountability
www.zubairkhalid.org ↗SK
SPEAKER 02
Dr. Salaar Khan
Adjunct Faculty, LUMS — Director, Learning & Development, CITY
Dr. Salaar Khan specializes in signal processing and artificial intelligence, with research at the intersection of advanced signal processing and AI-driven solutions. He holds a PhD from LUMS and an MS from Auburn University.
As Adjunct Faculty at LUMS, he teaches applied statistics, machine learning, data science, and Python, covering supervised and unsupervised learning, model evaluation, and hands-on training with popular ML libraries. In addition to his academic teaching, he co-delivers executive AI sessions that bridge technical implementation with leadership-level decision-making, and leads Learning and Development at CITY.
FOCUS AREASSignal processing & AI · Applied statistics & data science · Space-Time AI · Forecasting & predictive analysis
Join the masterclass ↗Learning outcomes
Knowledge that
compounds
Build and train deep neural networks from first principles by implementing forward propagation, loss computation, and backpropagation by hand, then verifying the results in code.
Design convolutional neural networks for vision tasks by implementing CNN architectures and articulating the architectural trade-offs against simpler models.
Implement generative architectures including autoencoders, GANs, and diffusion models, and compare their distinct approaches to data generation.
Explain and implement attention mechanisms and Transformers by tracing their evolution from RNNs and LSTMs and building a working attention mechanism.
Apply transfer learning to build and present an original capstone project, from architecture choice to a trained, evaluated model.
The 7-day plan
Seven sessions,
built layer by layer
Each session takes you one level deeper — from a single neural network to a complete, original project.
DAY 01OCT 02
Neural Network Foundations
Build and train your first neural network from scratch.
DAY 02OCT 05
Backpropagation Deep Dive
The math behind learning — by hand and in code.
DAY 03OCT 07
Convolutional Neural Networks
Image filters, CNN architecture, and ANN vs. CNN in practice.
DAY 04OCT 09
Generative AI Models
Autoencoders, GANs, and diffusion-based image generation.
DAY 05OCT 12
Sequence Models & Attention
From RNNs and LSTMs to the Transformer breakthrough.
DAY 06OCT 14
Transfer Learning
Pretrained models, fine-tuning, and domain adaptation.
DAY 07OCT 16
Capstone Project
Apply everything you've learned to a real-world project.
Who should attend
Built for those
ready to go deeper
01Developers and data scientists with working Python and machine learning experience who want a rigorous, hands-on deep dive.
02Students and researchers who've covered classical ML and are ready to build and train neural networks, CNNs, transformers, and generative models from the ground up.
03ML practitioners looking to move from applying pretrained models to understanding and implementing the architectures underneath them.
04Anyone with a technical foundation considering specializing in deep learning, computer vision, or applied AI research.
Organizers
Led by the CITY
at LUMS team
Mina Rafael Arif
Program Manager
Dr. Salaar Khan
Adjunct Faculty, LUMS — Director, Learning and Development, CITY
ENROLMENT OPEN / OCTOBER 2026
Ready to build
what's next?
Reserve your seat for the 7-session Deep Learning Masterclass at CITY at LUMS.
Enroll now ↗