AI & ML Track
AI & MLPath: From Zero to LLM
What is a Neuron?
Learn the foundational math of a perceptron, including weights, biases, and activation thresholds through an interactive demo.
Neural Networks
Deconstruct feedforward neural networks, backpropagation, and how information flows through layers in a step-by-step visual animation.
How Training Works
Visualize gradient descent, loss functions, and learning rates in real time with an interactive optimization graph.
Word Embeddings
Explore the vector space where words capture meaning. Map Word2Vec intuitions in an interactive 2D spatial coordinate visualization.
Attention Mechanism
Trace Query, Key, and Value interactions that power modern LLMs. Run step-by-step self-attention calculations on live token inputs.
The Transformer
Unpack the multi-head attention and encoder-decoder architecture that revolutionized modern Generative AI systems.
How LLMs Work
Learn what happens during pre-training, instruction fine-tuning, RLHF alignment, and active token generation cycles.
RAG Pipelines
Connect document chunking, semantic vector database retrieval, and model generation into an enterprise-ready pipeline.
Building AI Agents
Design autonomous AI agents utilizing tool execution, conversational memory buffers, planning steps, and orchestration routines.
Prompt Engineering
Master techniques like Chain-of-Thought, few-shot prompting, and structured schemas to generate highly reliable outputs.
Linear Regression
Understand the foundation of predictive algorithms. Fit lines of best fit and optimize slopes with gradient descent.
Logistic Regression
Learn binary classification models by mapping data to class probabilities using the mathematical sigmoid function.
Decision Trees
Walk through recursive binary splitting, calculating node entropy and Gini impurity metrics dynamically.
Random Forest
Construct ensemble classifiers that aggregate decision trees through bootstrap sampling and feature bagging.
K-Means Clustering
Cluster unlabeled data iteratively. Visualize centroid migration and class assignment steps in real-time.
Cosine Similarity
Calculate vector direction similarity to run semantic keyword search, context matching, and document rank algorithms.
Types of AI & AI Agents
A taxonomy of AI systems by capability and functionality, and a deep-dive into the six types of AI agents — from simple reflex to multi-agent systems.
Tokenization in NLP
How LLMs turn text into tokens via BPE, WordPiece, and SentencePiece — and why tokenization drives cost and context limits.
Fine-Tuning LLMs
Full fine-tuning vs. LoRA/QLoRA vs. instruction tuning, and when to fine-tune instead of prompting or RAG.
Evaluating LLMs
Benchmarks like MMLU and HumanEval, LLM-as-judge, perplexity, and hallucination metrics for measuring model quality.
Vector Databases Explained
Pinecone, Weaviate, and pgvector — approximate nearest neighbor search, HNSW indexing, and how vector databases power RAG.
Diffusion Models & Image Generation
Forward and reverse diffusion, denoising, and the intuition behind Stable Diffusion and DALL-E.
Generative Adversarial Networks (GANs)
The generator vs. discriminator adversarial game, mode collapse, and how GANs compare to diffusion models.
Reinforcement Learning Basics
Agent, environment, and reward — policy vs. value methods and the intuition behind Q-learning.
RLHF: Reinforcement Learning from Human Feedback
The reward model and PPO pipeline behind alignment — how ChatGPT and Claude were trained to follow instructions.
Convolutional Neural Networks (CNNs)
Convolution, pooling, and filters — how CNNs turn raw pixels into feature maps for image classification.
RNNs and LSTMs
Sequential processing, vanishing gradients, and LSTM gates — and why transformers eventually replaced them.
Overfitting & Regularization
Why models that ace training data fail in production, and the core fixes: L1/L2 regularization, dropout, and early stopping.
Hyperparameter Tuning
Grid search, random search, and Bayesian optimization — systematic strategies for finding good hyperparameters efficiently.
Support Vector Machines (SVM)
Max-margin classifiers and the kernel trick — and when SVMs still beat neural networks on small, high-dimensional data.
Naive Bayes Classifier
A probabilistic classifier built on a deliberately unrealistic independence assumption — still powering spam filters at scale.
PCA & Dimensionality Reduction
Compressing high-dimensional data into its most informative axes — eigenvectors, variance explained, and the curse of dimensionality.
Model Context Protocol (MCP)
Anthropic's open standard for connecting AI models to tools and data — the client-server architecture behind modern agentic AI.
Why LLMs Hallucinate
The structural causes of confident false outputs, and how RAG, grounding, and calibration reduce them.
Context Windows & Long Context
Tokens, quadratic attention cost, and how million-token context windows work — plus what causes context rot.
Multimodal AI (Vision + Language)
How vision-language models like GPT-4V, Claude, and Gemini tokenize images and reason across modalities.
AI Ethics & Bias in Machine Learning
How bias enters ML pipelines, fairness metrics, real-world cases, and responsible AI practices for engineers.