Kubernetes + AI Joining Forces in the Battle Against Cancer
Talkat Kubernetes AI Day Europe 2022 · on Monday, 2022-05-16· at 09:35· Pavilion 3, Room D | Level 2 | Central Forum·
summary
Uses stardust to segment nuclei, extracts 11x11 patches, compresses each to a 10‑dim latent with an autoencoder (training error 4.7 per pixel ≈ 12 gray levels), projects latents with UMAP, then k‑means into eight clusters; clusters reveal two promising internal-cell categories but show annoying border artifacts and remain unvalidated against clinical outcomes. Argues need for geographically distributed Kubernetes clusters near hospitals for petabyte storage, low‑latency networking and GPU scale; prototype analysis pipeline runs but clinical utility is not yet proven.
AI-written summary · about
chapters (18)
- 0:00Intro
- 2:33Goals
- 3:50Histopathologist
- 4:57White Rabbit
- 7:24Stardust
- 8:05Preprocessing
- 9:01Contrast
- 9:17Classification
- 9:52Autoencoder
- 12:50Examples
- 13:21Decoder
- 14:15Neural Network
- 16:16Results
- 19:22Archives of the Hospital
- 21:10Digitalization
- 23:56Storage
- 24:36Reference Links
- 25:13Questions
related talks
29mCancer diagnostics enabled by machines: finding and classifying single tumor....
27mGen-AI at Scale: Simplifying Orchestration of Healthcare Applications Across Multi-Cluster Environme
35mBuilding a Medical AI with Kubernetes and Kubeflow
38mAI in Kubernetes: How to Get Started?
38mKubernetes for AI Workloads – What Works and What… Doesn’t
21mFighting Cancer with AI
25mEmpowering AI-Driven Drug Discovery: Overcoming Challenges in Building a ML Platform on Kubernetes
22mGenerative AI Model Data Pre-Training on Kubernetes: A Use Case S...