Phase 1 — Computer Science, Algorithms & Software Architecture
In the AI era, anyone can generate a 50-line function. Very few can look at that function and immediately spot that it turns a 10,000-row API response into a 100,000,000-iteration CPU freeze, or that it tightly couples domain logic to a database driver.
Phase 1 builds your algorithmic intuition (Grokking Algorithms) and architectural discipline (Cosmic Python).
Modules & Lessons
01 — Complexity & Core Data Structures
01-complexity-and-arrays— Time/space complexity in production, dynamic arrays, amortized allocation, and slice pitfalls02-hash-tables-and-sets— Hash functions, collision resolution,dict/setinternals, and turning $O(n^2)$ nested loops into $O(n)$ hash joins03-stacks-queues-and-deques—collections.deque, monotonic queues, bounded ring buffers, and backpressure queues04-heaps-trees-and-tries—heapq, priority scheduling, Top-K queries, prefix trees (Tries), and URL routing tables
02 — Algorithmic Patterns for Backend Engineers
01-two-pointers-and-sliding-window— Fixed and dynamic sliding windows (implementing a production sliding-window rate limiter)02-binary-search-and-intervals— Binary search on monotonic predicates, log timestamp lookup, and time-window interval merging03-graphs-bfs-dfs-and-toposort— Adjacency lists, cycle detection, dependency resolution, and DAG workflow scheduling
03 — Software Architecture & Code Review
01-solid-and-cohesion— High cohesion, low coupling, Dependency Inversion, and avoiding speculative abstractions02-clean-and-hexagonal-architecture— Domain entities, Ports & Adapters, Repository pattern, and isolating business rules from frameworks03-code-review-and-ai-auditing— Systematic Code Review: catching race conditions, $O(n^2)$ loops, leaky abstractions, and AI hallucinations
Capstone Milestone: pulseforge/ v0.2
Upgrade PulseForge with a clean Hexagonal domain core, an in-memory sliding-window rate limiter per tenant, a bounded ring-buffer batcher, and a DAG alert-rule dependency evaluator.