PVPranjal Vermainblog.pvcodes.in·Aug 29 · 7 min readLocal AI Stack for Productive Small Language ModelsEvery day this week someone posted the same screenshot: a freshly pulled model chatting agreeably in a terminal, captioned "running a model locally." Fine. Now what? Because the gap between I got a mo00
PVPranjal Vermainblog.pvcodes.in·Aug 23 · 7 min readApache Iceberg Won the Format War - So Why Are We Rebuilding the Moat?For three years, every lakehouse architecture review started with the same tribal question: Iceberg, Delta Lake, or Hudi? In 2026, that question died. And the interesting part is what happened next — 20
PVPranjal Vermainblog.pvcodes.in·Aug 23 · 6 min readApache Iceberg: The Backbone of the Modern Open LakehouseThe lakehouse pitch is seductive: keep your data as open Parquet on cheap object storage, then let any engine - Spark for heavy ETL, Trino for dashboards, Athena for ad-hoc poking - query the same tab00
PVPranjal Vermainblog.pvcodes.in·Aug 12 · 13 min readSpark Performance Tuning Interview Prep GuideSpark interviewers stopped caring about API syntax years ago. At the senior and mid-level bar, the questions that separate candidates are almost all the same shape: "Your job is slow. Why, and what do00
PVPranjal Vermainblog.pvcodes.in·Aug 9 · 8 min readUnstructured data becomes a first-class citizen: Databricks FILE type and the multimodal lakehouseLast week Databricks announced the beta of FILE, a native column type that brings documents, images, audio, and video into your lakehouse tables with the same governance you already apply to structure00
