Catalogue

Cards
165
Records
486
Workloads
19
Corpus version
74847648e25a
Workload
Kind
Setup

46 of 486 records match; the first 20 are shown.

KindRecordOrganisationYearSetup
Failure

Traffic peak overloaded one datacenter's load balancers while a service-mesh sidecar failed to scale, and a client retry bug amplified authentication traffic

github-2026-load-balancer-sidecar-overload#f1

GitHub2026Multi-region
Decision

Partition catalog clustering to per-shop tasks and filter standalone stores deterministically via Liquid themes before calling LLMs

shopify-2026-catalog-clustering#d1

Shopify2026
Decision

Replace naive linkage-tree clustering with FAISS HNSW plus sparse UPGMA average linkage to form coherent chunks up to 200 products

shopify-2026-catalog-clustering#d2

Shopify2026
Decision

Use a two-stage propose-and-critique LLM flow with dynamic per-chunk Structured Output schemas and normalized product IDs

shopify-2026-catalog-clustering#d3

Shopify2026
Failure

Connection pools that recycled every 15 minutes synchronized into a thundering herd after a managed Postgres upgrade

clerk-2025-database-connection-recycling#f1

Clerk2025
Decision

Pair GPT-4 confirmation and MetaReflection hybrid prompts with expanded offline evaluation to curb LLM false positives

github-2025-copilot-secret-scanning#d1

GitHub2025
Decision

Share LLM capacity across push and full-history scans with dynamic rate-limit ranges and filepath exclusions

github-2025-copilot-secret-scanning#d2

GitHub2025
Decision

Validate prompt and filter changes via mirror testing on public-preview repositories before GA

github-2025-copilot-secret-scanning#d3

GitHub2025
Failure

Users who follow hundreds of thousands of accounts made hot shards that slowed the timelines sharing them

bluesky-2025-lossy-timelines#f1peak of 100,000 requests/s or more

Bluesky2025Sharded
Decision

Drop timeline writes with a probability set by how many accounts a user follows, accepting imperfect timelines

bluesky-2025-lossy-timelines#d1peak of 100,000 requests/s or more

Bluesky2025Sharded
Decision

Cache high-follow accounts in a Redis sorted set loaded into memory every 30 seconds instead of reading follow counts per write

bluesky-2025-lossy-timelines#d2peak of 100,000 requests/s or more

Bluesky2025Sharded
Decision

Use few-shot GPT-4-chat (gpt-4-32k) instead of fine-tuning davinci-002 or using the completion API

canva-2023-pir-summarizer#d1

Canva2023
Decision

Redact internal data pre-LLM, use soft 600-character prompt targets with retries, and track edits via Jira webhooks

canva-2023-pir-summarizer#d2

Canva2023
Decision

Run nearest-neighbor vector search in memory inside stateless Kubernetes services rather than a stateful database

spotify-2023-voyager-ann-search#d1

Spotify2023
Decision

Build Voyager as a new HNSW package to succeed Annoy instead of making breaking changes in hnswlib

spotify-2023-voyager-ann-search#d2

Spotify2023
Decision

Replace the Wide & Deep (XGBoost + DNN) fraud ensemble with a ResNeXt-inspired multi-branch pure DNN

stripe-2023-radar#d1

Stripe2023
Decision

Pair black-box DNN fraud scoring with merchant Risk Insights (backed by Elasticsearch) and internal feature-contribution tables

stripe-2023-radar#d2

Stripe2023
Decision

Move zone-level parallel ML training from a shared long-running Celery queue to per-application Ray clusters on EKS

instacart-2023-ray-distributed-ml#d1

Instacart2023
Decision

Isolate Python environments per Ray cluster behind a unified Griffin launcher API for dev and Airflow runs

instacart-2023-ray-distributed-ml#d2

Instacart2023
Decision

Restrict Venice to asynchronous batch and stream writes with background version swaps, server-side partial updates, and hybrid replay

linkedin-2022-venice#d1peak of 100,000 requests/s or more

LinkedIn2022Multi-region

26 more records. Sign in to see all 46.