If you're at the beginning of a BS degree in Computer Science, Software Engineering, Data Science, Machine Learning, or Artificial Intelligence at any university or affiliated college in Lahore or its vicinity, this is an excellent opportunity to build a strong foundation for your journey in this field.
Applied AI Engineering: From Probability to Generative Models & LLM Systems is a 6-month intensive, hands-on program designed to take you from core mathematical and algorithmic foundations to deploying state-of-the-art AI systems, including Generative Models, Transformers, and Large Language Models (LLMs).
Engage in mastering the essentials of linear algebra, calculus, probability, optimization, and neural networks, and then progress to building generative models (VAE, Diffusion), training transformers, and fine-tuning LLMs. Practice mathematical derivations, implement models in NumPy and PyTorch, benchmark performance, and deploy fully functional AI applications. By the end, you’ll be ready to build, optimize, and deploy AI systems, and work as an ML / Generative AI Engineer in real-world environments.
Math: Vector spaces, norms, dot product, matrix multiplication, linear transformations
NumPy Lab: Matrix multiplication from scratch, KNN classifier, 2D visualization
Torch Lab: GPU vs CPU tensors, batch operations & broadcasting
Mini Project: Implement KNN fully in NumPy
Math: Eigen decomposition, SVD, PCA derivation, low-rank approximation
NumPy Lab: PCA manually, image compression
Torch Lab: PCA on high-dimensional data, compare reconstruction errors
Math: Partial derivatives, gradients, Jacobian, chain rule, backprop idea
NumPy Lab: Compute gradients numerically, gradient descent, 3D loss visualization
Torch Lab: Autograd exploration, manual vs autograd comparison
Math: Convexity, SGD, Momentum, Adam, convergence
NumPy Lab: Implement optimizers, compare convergence rates
Torch Lab: Custom optimizer, learning rate scheduling
Random variables, PMF vs PDF, expectation, variance, covariance, Law of Large Numbers
NumPy Lab: Monte Carlo simulations
Torch Lab: torch.distributions exploration
Multivariate Gaussian, covariance matrix geometry, likelihood function
NumPy Lab: PDF implementation, contour plots, log-likelihood
Torch Lab: Fit Gaussian via gradient descent
MLE derivation, linear & logistic regression
NumPy Lab: Closed-form linear regression, logistic regression from scratch
Torch Lab: Build training loop, loss functions, checkpointing
Mixture models, EM algorithm, GMM as density estimator
NumPy Lab: EM implementation, clustering visualization
Torch Lab: Log-sum-exp trick, stable implementation
Backpropagation, activation functions, loss functions
NumPy & Torch Labs
Bias-variance, L1/L2, dropout, batch norm
Labs: Overfitting experiments, regularization comparison
Manual convolutions, Torch CNN for CIFAR/MNIST, train & evaluate
Autoregressive models, Normalizing flows, Energy-based models
ELBO derivation, KL divergence, reparameterization trick
NumPy & Torch Labs: Visualize latent space, full VAE implementation
Forward & reverse process, noise prediction objective
Torch Lab: Train simple diffusion, visualize denoising
Conditional VAE/diffusion, classifier-free guidance
FID score, likelihood vs perceptual quality, mode collapse
Scaled dot-product attention, multi-head attention
Encoder, decoder, positional encoding, mini transformer lab
Mixed precision, gradient clipping, checkpointing, experiment tracking
TorchScript, ONNX, FastAPI, Docker; deploy transformer as REST API
Causal LM, tokenization, scaling laws
Lab: Train tiny GPT from scratch
LoRA, QLoRA, instruction tuning, task-specific chatbot
Embeddings, vector DBs, hybrid search, FAISS lab
Quantization, vLLM, API deployment, monitoring lab
