Topics Directory

Master the foundational and advanced concepts of Machine Learning through structured modules. Prepare for technical interviews with high-density content.

↗
Easy

Linear Algebra for ML

Work through vectors, dot products, matrix shapes, projections and low-rank approximations. Then see the same ideas in embeddings, attention, PCA and LoRA.

3 Tradeoffs
⚄
Easy

Probability and Statistics for ML

Count outcomes, update beliefs, and tell an improvement from sampling noise. Work from probability to estimates, intervals and experiments.

3 Tradeoffs
∂
Easy

Calculus and Optimization

Trace a loss backward, take a gradient step, and understand when that step helps. Derivatives, chain rule, curvature and optimizers with numbers.

3 Tradeoffs
ⓘ
Easy

Information Theory for ML

Understand entropy, cross-entropy, KL divergence and perplexity through a small prediction problem, then connect them to training and compression.

3 Tradeoffs
≈
Medium

Numerical Computing for ML

Make the math survive real hardware: floating point, stable softmax, conditioning, gradient checks and mixed precision.

3 Tradeoffs
📐
Easy

ML Fundamentals

Build a trustworthy train/validation workflow: splits, bias and variance, regularization, calibration and production diagnosis.

5 Tradeoffs
🤖
Easy

Classical ML

Choose an evaluation target, fit interpretable baselines, and reason through kernels, tree ensembles, clustering, and static word representations.

5 Tradeoffs
🧠
Easy

Deep Learning Basics

Trace neural-network updates with worked derivatives and tensor shapes, then reason about activations, normalization, optimizers, losses, regularization, and numerical failures.

5 Tradeoffs
🏛️
Medium

Pre-Transformer Architectures

Work through convolution geometry, recurrent states, gated memory, residual paths, and selective state-space models to understand architecture and deployment tradeoffs.

5 Tradeoffs
🧠
Medium

Transformers & Attention

Calculate attention outputs, track tensor shapes and masks, then connect rotary positions, transformer blocks, exact attention kernels, and KV caches to real costs.

4 Tradeoffs
🔍
Medium

Embeddings & Retrieval

Connect learned vector geometry to contrastive training, approximate search, hybrid retrieval, filtering, evaluation, and safe index updates with worked examples.

4 Tradeoffs
📊
Medium

Evaluation & Benchmarking

Choose a decision-relevant metric, compare models on the same cases, and separate real improvements from noise or measurement errors.

4 Tradeoffs
🔧
Medium

Fine-tuning Strategies

Adapting pre-trained models to specific tasks with full fine-tuning, LoRA, QLoRA, and other PEFT methods

3 Tradeoffs
💬
Medium

Retrieval-Augmented Generation

Grounding LLM outputs with external knowledge via retrieval pipelines

7 Tradeoffs
🤖
Hard

Agentic AI

Autonomous AI agents that plan, use tools, and execute multi-step tasks

4 Tradeoffs
🪢
Hard

Harness Engineering for AI Agents

Design the state, tools, checks and recovery loop that turn a model’s proposed actions into inspectable work.

4 Tradeoffs
🎯
Hard

Reinforcement Learning Basics

MDPs, value functions, policy gradients, actor-critic, PPO, and the bridge to RLHF/GRPO for LLMs

4 Tradeoffs
🎯
Hard

RLHF & Alignment

Aligning LLMs with human preferences using RLHF, DPO, and modern preference optimization methods

3 Tradeoffs
🧪
Hard

Test-Time Compute & Reasoning Models

Allocating inference work across longer solutions, multiple candidates, verification, and search under quality, cost, and latency constraints

3 Tradeoffs
🛡️
Hard

Safety, Alignment & Red Teaming

Threat models, trust boundaries, red teaming, and measurement of harmful compliance and benign refusal

4 Tradeoffs
🧩
Hard

Distributed Training & Parallelism

Work out what lives on each GPU, which tensors cross the network, and why a model can fit yet still train slowly.

3 Tradeoffs
⚡
Hard

Model Quantization

Calculate what low precision saves, follow rounding and clipping by hand, and test whether those savings survive in a real serving workload.

4 Tradeoffs
🚀
Hard

Model Serving & Inference Optimization

Follow a request from queue to first token, size its KV cache, and build a serving plan from measured latency and capacity.

4 Tradeoffs
🧩
Hard

Mixture of Experts

Trace sparse routing, expert capacity, balancing losses, and parameter budgets, then connect them to distributed execution and serving measurements.

3 Tradeoffs
👁️
Hard

Multimodal Models & VLMs

Trace images into language-model inputs, calculate visual token costs, and test whether answers are supported by what the model can see.

4 Tradeoffs
🎨
Hard

GANs, Diffusion & Flow Matching

Follow adversarial learning, denoising and flow matching through small calculations, then compare their sampling and deployment costs.

4 Tradeoffs
🌍
Hard

World Models

Learn to predict action-dependent futures, plan with a small dynamics model, and test whether imagined rollouts remain useful in the real environment.

5 Tradeoffs
🤖
Hard

Vision-Language-Action Models

Connect images and instructions to robot commands, calculate action precision and timing, and evaluate closed-loop behavior beyond a demo.

5 Tradeoffs