Day 5
summary
Front‑end talk shows a concrete workflow: measure with Lighthouse/WebPageTest, set asset size budgets and CI hooks, remove a full Lodash import (swapping a tiny vanilla sort) and recover huge vendor savings — real teams see revenue jump from ~0.3s latency cuts. GraphQL Mesh converts OpenAPI/grpc/SQL/Neo4j to GraphQL, stitches or federates sources (Microsoft Graph converted in minutes), and publishes reusable mesh modules. Deep learning rules give practical wins: tf.data + prefetch/autotune, mixed precision (FP16 ≈ 2–3×), batch sizes in multiples of 8/64, DALI/JPEG decode on GPU, 8‑bit quant retains ~98% accuracy. Conversational demo streams browser audio via WebRTC/RecordRTC + socket.io-stream → Speech‑to‑Text → Dialogflow → Translate → WaveNet TTS and deploys on App Engine; it works end‑to‑end.
AI-written summary · about
chapters (7)
- 0:00Welcome to day five of WeAreDevelopers Live Week!
- 23:23Frontend performance testing in practice – Jonas Kröger
- 1:36:10GraphQL Mesh – Why GraphQL between services is the worst idea and the best idea at the same time! – Uri Goldshtein
- 3:07:17Quantum Computing: Why we should and how we can better compute the future – Alexandra Waldherr
- 4:20:3630 golden rules of Deep Learning performance – Anirudh Koul
- 5:36:38Raise your voice! – Lee Boonstra
- 6:24:11Closing session of the first WeAreDevelopers Live Week
