RecallRun/Learn
11 stacks · diagram-first · plain English + engine level · every course ships a project

Pick a stack. Learn it properly.

Not tutorials that list functions. Each course builds the mental model of the machine first — then the API, then a hands-on project with deliberately messy data. Every concept gets an analogy, a hand-drawn diagram, runnable code, the gotchas, and a checkpoint quiz.

Core Python

From zero to confident
1

Start from nothing: values and types, strings, control flow, collections, functions, modules, files, exceptions, classes and testing.

  • How Python runs your code, and a proper setup
  • Strings, lists, dicts and sets, the Pythonic way
  • Functions, scope, modules, files and exceptions
  • Classes, inheritance, debugging and pytest
21 lessons+ project
Open Core Python →
not started0%

Advanced Python

How the language works
2

Names vs objects, CPython internals, decorators, generators, descriptors, metaclasses, typing, the GIL and asyncio.

  • The object model that explains every "weird" bug
  • Decorators, iterators, dunder methods, descriptors
  • Typing, exceptions, context managers
  • GIL, threads vs processes, asyncio, profiling
21 lessons+ project
Open Advanced Python →
not started0%

DSA in Python

Structures & algorithms
3

Big-O measured on real timings, every core structure drawn and implemented, and the patterns behind interview problems.

  • What Python's list/dict/set really cost
  • Two pointers, sliding window, hashing, binary search
  • Trees, heaps, tries, graphs, shortest paths
  • Backtracking, greedy, dynamic programming
25 lessons+ project
Open DSA →
not started0%

SQL

The language of data
4

Engine-agnostic SQL taught by execution order — joins, subqueries, CTEs, window functions, modelling, indexes and EXPLAIN.

  • The logical order that explains SQL's errors
  • Joins without the fan-out and NULL traps
  • Window functions, dates, strings, JSON
  • Normalization, transactions, indexes
28 lessons+ project
Open SQL →
not started0%

MySQL

The server & InnoDB
5

From the query down to the disk: server architecture, the clustered index, MVCC, locking, EXPLAIN, backups and replication.

  • Buffer pool, redo/undo logs, crash safety
  • Composite indexes & the leftmost-prefix rule
  • Isolation levels, gap locks, deadlocks
  • Tuning, backup/PITR, replication, security
22 lessons+ project
Open MySQL →
not started0%

FastAPI

Production web APIs
6

ASGI and Pydantic v2 internals, dependency injection, async done right, security, testing, observability and deployment at senior depth.

  • Request lifecycle, validation and dependencies
  • Async concurrency, streaming, SSE and WebSockets
  • Async SQLAlchemy, JWT auth, OWASP API security
  • Testing, observability, performance, deployment
25 lessons+ project
Open FastAPI →
not started0%

System Design

1 user → 1 million
7

One product's journey from a single server to a million users: every load balancer, cache, shard and region added only when something breaks.

  • Estimation, queueing and the latency numbers, measured
  • Load balancing, caching, CDNs, replication, queues
  • Sharding, consistent hashing, rate limits, resilience
  • SLOs, feeds, multi-region, security, cost, interviews
24 lessons+ project
Open System Design →
not started0%

PySpark

Distributed processing
8

Driver, executors, shuffles and Catalyst — why a job is slow and how to fix it, from your first SparkSession to production.

  • Cluster architecture & the Py4J bridge
  • DataFrames, joins, windows, UDFs
  • Shuffle, memory model, Spark UI, skew, tuning
  • Streaming, Delta Lake, MLlib, production
30 lessons+ project
Open PySpark →
not started0%

Databricks

The lakehouse platform
9

Everything around Spark: Unity Catalog, Auto Loader, declarative pipelines, jobs, Databricks SQL, MLflow, cost and CI/CD.

  • Control plane vs compute plane, serverless
  • Unity Catalog governance & Delta features
  • Ingestion, medallion, pipelines, jobs
  • Performance, cost, bundles & CI/CD
18 lessons+ project
Open Databricks →
not started0%

OpenAI SDK

LLM calls in Python
10

From the first request to production: the Responses API, streaming, structured outputs, tools, embeddings, retries, cost and agents.

  • The request loop every LLM feature is built on
  • Pydantic-typed outputs and function calling
  • Async, errors, rate limits, caching, Batch API
  • The Agents SDK and testing without the network
20 lessons+ project
Open OpenAI SDK →
not started0%

AI Engineering

LLMs · RAG · MCP · observability
11

How LLMs work, retrieval-augmented generation, agents and the Model Context Protocol, then OpenTelemetry, Langfuse and Grafana.

  • Tokens, attention, sampling, and why models hallucinate
  • Embeddings, vector search, chunking, RAG evaluation
  • Agents, MCP servers and clients in Python
  • Tracing, metrics, evals, guardrails and cost control
21 lessons+ project
Open AI Engineering →
not started0%

How every course is built

🧠 Analogy first

Every concept opens with a kitchen, a library or a sorting office — no jargon until the idea has landed.

⚙️ Then engine truth

What the interpreter, planner, storage layer or scheduler is actually doing, at senior-engineer depth.

📊 A diagram per idea

Hand-drawn SVG that adapts to light and dark mode and stays readable on a phone.

💻 Code that runs

Complete, copy-pasteable snippets. Where they could be executed (Python, DSA, SQLite, FastAPI, the AI projects), the outputs shown are real.

🚨 Gotchas marked

The mistakes that cost real teams real money, with the fix next to them.

✅ Checkpoint quizzes

Two to four questions per lesson with explanations, so you can prove you understood it.

Every course ships a project

CourseProjectRuns with
Core PythonAn expense-tracker CLI: a messy 600-row bank export cleaned, categorised and reported, as a package of small modules with 53 testsPython 3.12+, standard library only
Advanced PythonA small task-queue library: decorators, descriptors, generators, asyncio, typing, pytestPython 3.12+
DSAA route planner over 26 South Indian towns: Dijkstra/A* on your own heap, LRU cache, trie autocomplete, 15 testsPython, no dependencies
SQL5-table retail database, 4,007 messy rows, 30 graded challenges — every solution verified by executionSQLite (zero install), Postgres, MySQL
MySQLMake a slow database fast with EXPLAIN ANALYZE, provoke a deadlock, point-in-time restoreMySQL 8 in Docker
FastAPIA production-grade orders API: JWT roles, idempotency keys, ETags, live SSE status, a transactional outbox for signed webhooks, rate limits, 51 async testsPython + SQLite (zero setup), PostgreSQL via Docker
System DesignScale Lab: a URL shortener taken through six architecture stages under real load (index, click queue, cache, load balancer, shards), a live failover, and a capacity plan for a million usersPython, no services to install
PySparkRetail analytics over CSV + JSON + Parquet: clean, join, window, write partitionedLocal Spark or Databricks
DatabricksEnd-to-end lakehouse: Auto Loader → declarative pipeline → jobs → dashboardDatabricks Free Edition
OpenAI SDKA retail analyst assistant: SQL tool calling, structured answers, streaming, embeddings search, cost tracking, tests with a fake clientPython + OpenAI API (tests run offline)
AI EngineeringAn observable docs copilot: RAG over this site, exposed as an MCP server, traced with OpenTelemetry → Grafana, evaluated in LangfusePython + Docker

Also here

⚡ PySpark one-page tour →

A fast single-page overview of the Spark mental model — a 20-minute refresher before an interview.

📚 The library

Eleven courses, each with runnable examples, quizzes and a hands-on project.

Progress is stored in your browser only. Nothing is uploaded anywhere.