Featured Case Studies
Real production work with measurable outcomes
Scale-to-Zero GPU at 116 GPUs
Karpenter + KEDA HTTP add-on + metrics-api across 5 model types (LLM, embedding, OCR, vision, multimodal). $0 idle, 116 GPUs including 55× RTX PRO 6000 Blackwell.
NVIDIA Container Toolkit Regression Fix
Found, reproduced & got fixed upstream: regression in NVIDIA Container Toolkit 1.20.0 breaking GPU detection on EKS AL2023 NVIDIA AMI. Filed #2026, NVIDIA merged fix in #2028 (1.20.1).
Cutting GPU Cold Starts: 54% Faster
Benchmarked NVMe, Bottlerocket, SOCI lazy loading, parallel pulls, FastImagePull, EBS throughput. Best: 43s (54% faster) with AL2023 FastImagePull + 600 MiB/s EBS.
Documentation Guides
Comprehensive, step-by-step references
Running GPUs in 2026
6 provider guides: Brev, Verda, E2E, AWS, GCP, Local. Compare GPUs, pricing, setup, and use cases.
और जानें →Cost Optimization Playbook
Systematic AWS cost reduction: Compute, Database, Storage, Networking, Monitoring. Not tips — a framework.
और जानें →