Case Studies
Case Studies
Long-form analyses of production ML systems — architectures, post-mortems, and head-to-head comparisons. Where the engineering has documented consequences, we try to document the engineering.
-
Inside PlateLens's Calorie-Accuracy Claim: A Technical Replication
The three-stage CV pipeline (ViT-L/16, ZoeDepth, USDA-aligned DB) behind PlateLens's accuracy, benchmarked against MyFitnessPal and Cronometer.
By Dr. Marcus Brennan ·
-
Anatomy of a production ML failure: Zillow's iBuy collapse
What the engineering team got wrong, what was structurally hard, and how the public post-mortem reads two years later.
By Dr. Nadia Volkov ·
-
CRDTs in production: lessons from Figma's multiplayer engine
The original 2023 teardown of Figma's per-property CRDT model and its server-authoritative fan-out.
By Priya Ramachandran ·
-
Discord's architecture: why they're migrating from Elixir to Rust
The technical reasons behind Discord's gradual shift from Elixir to Rust in its core messaging path.
By Priya Ramachandran ·
-
Plaid's bank integration API: a system design study
How Plaid structured its bank-integration API around reliability, multi-tenant key management, and rate-limit negotiation.
By Priya Ramachandran ·