Learning Paths
Not sure where to start? A path is a hand-ordered track of subjects — learn, practise, and prove it with a quiz, one step at a time.
Claude Code Mastery
A guided path through Anthropic's agentic coding CLI — from your first session to configuring it well for real projects (CLAUDE.md, memory, settings, permissions and hooks), working across its surfaces (IDE extensions, desktop, web, mobile, Slack), choosing models and managing cost and context, using its higher-autonomy planning and workflow modes, extending it with skills, slash commands and plugins, orchestrating MCP servers, subagents, dynamic workflows and agent teams, running it in git workflows, sandboxes, schedules, cloud routines and CI/headless mode, and the working practices that keep agentic sessions reliable and cheap — capped by an end-to-end case study.
AI Engineer Interview
What it takes to design and ship production LLM applications — and to explain the reasoning in an interview. Starts with the model fundamentals (transformers, tokenization, the full pretraining-to-chatbot pipeline, prompting, fine-tuning) and reasoning models (inference-time scaling, training reasoning models with RLVR and reward models), then context engineering and multimodal models, retrieval-augmented generation end to end (architecture, chunking and embeddings, vector search, evaluation), agents and tool use (workflow design patterns, the agentic loop, memory, multi-agent orchestration, MCP, agent evaluation), and production concerns (guardrails and prompt-injection defense, cost and latency engineering, observability, local and open-weight deployment). Surveys the framework landscape, then finishes with five full worked designs: a customer-support assistant, a web-search agent, a deep-research agent, a Claude Code-shaped coding agent, and an RLHF data and training platform.
Prompt & Context Engineering
Prompting treated as an engineering discipline, not typing nicely — from tokens and context windows, through few-shot/chain-of-thought and the advanced technique landscape, to engineering the full context window (selection, ordering, compaction, isolation), memory, injection defense, evaluation and versioning, cost/caching optimization, and context engineering for long-running agents. Closes with two case studies: designing the context pipeline for a long-context document assistant, and migrating a production prompt suite across a model deprecation.
Data Science Foundations
The full analytical toolkit a data scientist is expected to own — probability and statistics, hypothesis testing and A/B experiments, regression and regularization (L1/L2), model evaluation and validation, feature engineering, tree ensembles, unsupervised learning, SQL, causal inference, and time series. Free to start, ordered so each subject builds on the last.
System Design Interview
The complete sequence to prepare for a system design interview: the four-step interview framework and capacity estimation, then every building block interviewers expect you to reason about — load balancing, caching, consistent hashing, replication and sharding, SQL vs NoSQL data modelling, the CAP theorem, indexing, message queues, rate limiting, API design, distributed transactions and consensus, and reliability patterns — finished with three classic end-to-end designs: a URL shortener, a news feed, and a chat system.
ML System Design Interview
What an ML engineer must know to take a model to production — and to explain it in an interview. Starts with the interview framework and the generic system-design vocabulary, then covers data pipelines and feature stores, training and experimentation, serving and deployment, monitoring and drift, ranking and recommendation architecture, LLM applications, and multimodal vision-language models. Finishes with six end-to-end business cases modelled on real products: video recommendation, ad click prediction, payment fraud detection, search ranking, virtual try-on, and visual anomaly detection in manufacturing.
UI/UX Design Interview
What it takes to design well and defend the reasoning behind every decision in a UI/UX or product design interview. Starts with the UX/UI distinction and the process frameworks (Double Diamond, Design Thinking, human-centered design) interviewers expect you to adapt, not recite, then covers user research methods, personas/journey maps/JTBD, information architecture, interaction design principles (Nielsen's heuristics, Gestalt, Fitts's and Hick's Laws), wireframing and prototyping, visual design fundamentals, design systems and dev handoff, usability testing and evaluation, accessibility and inclusive design, responsive and mobile design, UX writing and microcopy, tying design to metrics and product outcomes, giving critique and presenting a portfolio, and design ethics and dark patterns. Finishes with two full worked whiteboard-style case studies: redesigning a checkout flow and designing a mobile app from scratch.
SQL & Data Engineering
The modern data engineering interview loop, end to end — SQL depth (execution order, window functions, query optimization), dimensional and normalized data modeling, warehouses and lakehouses, file formats and storage layout, ETL/ELT pipeline design, orchestration with Airflow, analytics engineering with dbt, Spark architecture and performance tuning, streaming fundamentals with Kafka, and data quality/observability in production. Closes with three case studies: designing a batch analytics platform end to end, a streaming event pipeline, and a worked gauntlet of hard SQL interview problems.
Algorithm Interview (LeetCode Interview)
The classic coding-interview pattern curriculum, end to end — the twenty reusable techniques that cover the overwhelming majority of LeetCode-style interview questions. Starts with the array/string fundamentals (two pointers, hash maps and sets, sliding windows), moves through linked-list and pointer techniques (fast and slow pointers), searching and ordering (binary search, sort and search), the core data-structure toolkit (stacks, heaps, intervals, prefix sums, trees, tries, graphs), then the two algorithmic paradigms that trip candidates up most (backtracking, dynamic programming), and closes with greedy algorithms, bit manipulation, and math and geometry. Each subject pairs a pattern write-up with hands-on problems and a quiz gating subject completion.
Reinforcement Learning & Long-term Optimization
From MDPs and bandits to policies that optimize what actually matters — cumulative, delayed user value instead of today's clicks. Covers the full method ladder (bandits, value-based, policy gradients, PPO, model-based), then the production reality: reward design, off-policy evaluation, safe exploration, and serving. Closes with three end-to-end cases: notification timing for a messaging product, a recommender optimizing long-term engagement, and an RLHF data and training platform.
MLOps & Production ML
The engineering spine around a model: reproducible environments, CI/CD for model and data code, experiment tracking and registries, lifecycle governance, progressive rollout, and observability once it's live. Mostly assembled from existing production subjects plus seven new ones, ending with a batch-platform case, one churn model followed from notebook to governed, monitored production, and a visual anomaly detection case covering per-SKU model zoos and edge deployment.
Generative Vision & Image AI System Design
From how images are generated — GANs, autoregressive tokenizers, diffusion, personalization, video — to how image AI actually ships as a business: virtual try-on, generative fill, catalog photography, super-resolution, synthetic training data, factory defect detection and deepfake detection. Each case study ends with how the real product solved scale, what it costs, and what an interviewer will push on.
Sign up free to track your progress across every path.
Get started free →