Skip to main content

Now (October 2026)

Inspired by Derek Sivers' /now movement

🎯 Current Focus​

Building: A GPU inference platform that scales to $0 idle — Karpenter + KEDA HTTP add-on + metrics-api scalers across 5 model types (LLM, embedding, OCR, vision, multimodal). Running on Blackwell (B200/B300), L40S, and L4 spot fleets.

Optimizing: Post-credits cloud cost strategy — splitting bills into baseline (commit) vs variable (on-demand). 1-year no-upfront Compute Savings Plan + targeted RIs for databases. ODCR for business-hours GPU capacity.

Learning: NVFP4 quantization on Blackwell, FlashInfer attention backends, vLLM 0.6+ continuous batching improvements, and the Karpenter 1.0 GA release.

Writing: Finishing the "Running GPUs in 2026" doc series (6 providers), EKS Security Hardening guide (5 layers), and AWS Account Setup playbook (5 guides).

🔧 Active Experiments​

  • SOCI parallel-pull + FastImagePull for sub-50s GPU cold starts
  • Value-based GPU taints (nvidia.com/gpu=workload-x) for perfect NodePool isolation
  • vmagent + vmauth for edge device metrics (20 GB disk queue > Alloy's 8-hour WAL)
  • Business-hours ODCR via EventBridge + Lambda for g6 capacity guarantees

📚 Reading​

  • Designing Data-Intensive Applications (re-read, Chapter 7: Transactions)
  • NVIDIA Blackwell architecture whitepapers
  • The Staff Engineer's Path by Tanya Reilly

🎤 Speaking​

📫 Open To​

  • Senior/Staff/Principal roles: DevOps, Platform, MLOps, Cloud Architecture
  • Consulting: GPU inference platforms, EKS hardening, cloud cost optimization
  • Conversations: GPU scheduling, FinOps, Kubernetes security, observability

Last updated: October 2026. This page changes monthly.