01 / FIELD NOTES
SQL foundations 7 min readOne table. Three answers. Conditional aggregation in SQL.
Build a payment summary with CASE, SUM and COALESCE—and verify its meaning with a tiny dataset.
Foundation
THE CODEYOURSELF FIELD GUIDEUnderstand the system.
Understand the system.
Then build it.
Clear explanations. Small, worked examples. Engineering ideas you can take into your next project.
A SMALL MODEL, A CLEARER ANSWER01 — 03
A good questionWhat are we measuring?→A small modelMake assumptions visible.→A better decisionTest what changes.
Read → reason → practiceBuilt for curious engineers
FOLLOW YOUR CURIOSITY
Go a little deeper in one direction.
LESS THEORY IN A VACUUM
Good ideas, made practical.
02 / FIELD NOTES
System design 6 min readA 90% cache hit ratio does not remove 90% of all traffic.
Separate reads from writes to calculate backend pressure, then test the assumptions that make the number useful.
Foundation
03 / FIELD NOTES
System design 6 min readA queue buys time. It does not create processing capacity.
Calculate backlog, drain time and overflow with a small burst model—and know what the model leaves out.
Foundation
04 / FIELD NOTES
AI engineering 7 min readYour LLM has two budgets: throughput and context.
Estimate sustainable request rate from prompt size, output size and repeated calls without confusing it with the context window.
Intermediate
05 / FIELD NOTES
Snowflake 6 min readSize a Snowflake warehouse around the workload, not a guess.
Separate query execution, concurrency and cost questions before changing compute size.
Intermediate
06 / FIELD NOTES
Databricks 6 min readA streaming checkpoint is progress, not your whole recovery plan.
Reason about restarts, replay and side effects before claiming reliable streaming.
Intermediate
07 / FIELD NOTES
API engineering 7 min readA timeout is not a failed operation. Design safer API retries.
Trace a duplicate order through request identity, parameter matching and a recovery ledger.
Intermediate
08 / FIELD NOTES
System design 6 min readAn error spike needs a denominator. Reason with burn rate.
Calculate a request-based error budget and distinguish a warning signal from a paging policy.
Intermediate
09 / FIELD NOTES
Databricks 7 min readA clean table needs a contract. Design useful quality gates.
Define grain, rejected-row reasons and reconciliation before calling a dataset business-ready.
Intermediate
FROM UNDERSTANDING TO DOING
Make the next step a small one.
Create a Free account for six coding challenges and five starter labs. Advanced labs and Architecture Studio are Pro features.
Start free practiceSave your own workPractice with browser fixturesClear Free & Pro accessSQL practice uses local DuckDB fixtures. System-design simulations model assumptions; they do not provision real cloud services.