Learning paths

Pick a reading path.

Curated routes through the Decipher Engineering library, ordered so each article builds the next piece of the mental model.

Intermediate23 lessons

LLM Inference Systems

Understand the memory, scheduling, decoding, and hardware tradeoffs that make LLM serving fast or painfully expensive.

135 minllm-inference-systems
  1. What Problem Did Language Models Try to Solve?
  2. What Came Before Transformers?
  3. Why N-Gram Models Hit a Wall
  4. Why RNNs Process Text Sequentially
  5. 19 more lessons
Open path
Advanced3 lessons

Serving Performance

A short path through the runtime decisions that turn model capability into production latency and throughput.

17 minserving-performance
  1. The Memory Trick Behind Fast LLM Serving
  2. How Speculative Decoding Makes LLMs Feel Faster
  3. 17,000 Tokens per Second, Burned Into Silicon
Open path
Foundational2 lessons

Data Systems Internals

Build the mental model for why distributed and analytical data systems need coordination, metadata, and storage layout choices.

11 mindata-systems-internals
  1. Why Kafka Needed ZooKeeper
  2. Why Columnar Databases Feel Like Cheating
Open path
Menu