Industry Case Clinic

About this course

A weekly clinic that dissects one real, publicly documented industry system or incident per session: how chat, payments, feeds, search and generative-AI products were built at scale, and what major outages at cloud providers and platforms teach about failure. Learners read the primary source in advance, reconstruct the decisions and the causal chain, debate the options the team had, and transfer the lessons into decision records for their own systems.

For: architect, software engineer, tech leader, ml practitioner.

Learning outcomes

On successful completion you will be able to:

1. 1Analyse how large-scale production systems in messaging, payments, feeds, search and generative AI are architected, and identify the forces that shaped them.
2. 2Reconstruct the causal chain of a major production incident from a public postmortem, separating the trigger, contributing factors and latent conditions.
3. 3Evaluate the options a team had at each decision point and the evidence for the outcome claimed, distinguishing what the source reports from inference.
4. 4Transfer lessons from an industry case into concrete architecture recommendations and decision records for a different context.
5. 5Lead a blameless case discussion with peers, defending a position while revising it in response to evidence.

Modules and lessons

1. Week 1Messaging at scale⌄

   • Storing trillions of messages at Discordcase study
2. Week 2Payments correctness⌄

   • Idempotency and safe retries in payments APIscase study
3. Week 3Feeds and recommendations⌄

   • The multi-stage ranking funnel behind Instagram Explorecase study
4. Week 4Search ranking⌄

   • Moving search ranking from trees to deep learning at Airbnbcase study
5. Week 5Generative-AI products⌄

   • Building a generative-AI product at LinkedIncase study
6. Week 6Cloud-scale failure⌄

   • The 2020 Amazon Kinesis outage in US-EAST-1case study
7. Week 7Network and dependency failure⌄

   • When the backbone disappeared - Meta, October 2021case study
8. Week 8Data loss and operational discipline⌄

   • The GitLab.com database outage of January 2017case study

Assessment

• Case analysis (50%)

  Two written case analyses (about 1,500 words each) of documented cases outside the clinic list, one system design and one incident.

• Design review (30%)

  Transfer memo with three ADRs applying lessons from one clinic case to the learner's own system.

• Peer review (20%)

  Facilitation of one clinic discussion, assessed by the mentor and by peers against a rubric.

Pass mark 60%. Grades: Distinction, Merit, Pass.