Scaling Personalization: DevOps at Stitch Fix

on YouTube ↗32m

" summary

Stitch Fix runs personalized recommendations by scoring every client × inventory pair daily with machine learning, combines those scores with 3500 stylists for curation, and runs a tech stack of microservices, event-driven pipelines, and CI/CD with 125 engineers and 80 data scientists. Engineering scales via small cross‑functional two‑pizza teams, TDD, A/B experiments (roughly 1/3 win, 1/3 neutral, 1/3 lose), cloud‑only ops and integration of 50 third‑party services, enabling rapid iteration and tangible revenue and headcount growth.

AI-written summary · about

" chapters (16)

1. 0:00Intro
2. 0:19Background
3. 1:24Stitch Fix
4. 2:29Personalized Recommendations
5. 3:39Humans + Data Science
6. 6:28Scaling Personalization
7. 6:50Conway's Law
8. 7:50Small "Service" Teams
9. 13:28What Problem Are You Trying to Solve?
10. 14:24Buy, Not Build
11. 18:48Experimental Discipline
12. 23:07Evolution to Microservices
13. 26:38Quality Discipline
14. 27:15Test-Driven Development
15. 28:23Optimizing Developer Effort
16. 29:51Build It Right (Enough) The First Time

" related talks

• 25mPersonalization and Retail: Lessons from Stitch FixBrad Klingenberg · Computational Theories of the Brain 2018
• 40mEnabling simple ML ops at Stitch FixElijah ben Izzy · Data Con LA 2021
• 34mAugmenting Human Decision Making with Data ScienceKelsey Pedersen · PyCon 2018
• 20mBrad Klingenberg, Director of Styling Algorithms, Stitch FixBrad Klingenberg · MLconf SF 2015
• 02mDBTB IN Eric Colson rEric Colson · Data By the Bay 2016
• 18mSynthesizing human and machine capabilitiesEric Colson · Data By the Bay 2016
• 39mDecision Making Based on Machine Learning at OutfitteryJesper Richter-Reichhelm · code.talks 2017
• 42mAn Algorithm of StyleEd Snelson · PyData London 2017