Available for collaboration
Measured local AI. Production-minded infrastructure.
I'm Alfonso. Local AI is my obsession: I benchmark local models on my own DGX Spark and publish field notes from building secure, cloud-native infrastructure.
Tools and technologies I work with: Local AI, DGX Spark, Kubernetes, GitOps, + 6 more
Latest writing
All articlesOne DGX Spark, four ways to serve Qwen3.8-Flash-Next
LLMQwen3.8-Flash-NextDGX Spark
I served Qwen3.8-Flash-Next four ways on one DGX Spark. Answer quality is close. Speed, memory and structured output are not.
Measured on one DGX SparkDecode, geometric mean: hibrid48 54.3 t/s, TensorFold 51.6, fast 45.8, quality 44.4Real-world speed, measured by me. No vendor marketing numbers.- Why I'm Betting on Local AI (and How You Can Start)
- DeepSeek V4 Flash vs GLM 5.3 Flash: Who Wins on a Single Spark?
- DeepSeek V4 Flash on a Single DGX Spark
- Running Laguna S 2.1 on a DGX Spark
