Hello, I'm
I write about technology, ideas, and things I'm learning. Welcome to my corner of the internet.
50 harder-than-recall questions per series, scored as you go. Pick one.
Chart of the day
Apple vs Google maps
The UX on navigation is better for Apple Maps on iPhone. Google Maps just never got that smooth in its turn-by-turn navigation on CarPlay. It has on Android Auto. Something probably 99.9% of people wouldn't care about, but somehow it bothers me.
Am willing to sacrifice real-time traffic data and times for the UX. So much so that I risked using Apple Maps on a near 1,500 km trip, and somehow did end up using Google Maps for 5% of the time. In India. That's a decent win for Apple Maps.
Of course, Apple Maps is way off when it comes to listed places.
As part of the Lab 1 in MIT's distributed Systems course, this github repo is to explain the workings of MapReduce jobs with ample pretext of fundamentals and motivation behaind distributed system workflows covered in the blog link in this card.
The hands-on labs behind the second half of my Agentic AI series: Strands agents deployed to AWS on Lambda, ECS Fargate and Bedrock AgentCore, then traced with OpenTelemetry into CloudWatch, Langfuse and LangSmith, scored with AgentCore Evaluations, and limited with Bedrock Guardrails and Cedar policies. The Blog link opens the write-up of the deployment labs.
Building my own shell program using Rust programming language.
A natural language task scheduler that uses AI to parse plain English into structured reminders and to-dos.
This very site: a Next.js 16 and Tailwind CSS 4 blog and portfolio where every post is an MDX file in the repo. Posts are organised into sections and ordered series, each with a table of contents, and a Listen to this post player reads them aloud using the browser's text-to-speech. It has a vanilla-and-olive theme with light and dark modes and is deployed on Vercel. The link opens the build log that walks through the code.
A trip planner for a group of friends that grows lesson by lesson: SQLite, embeddings and ChromaDB retrieval, an LLM chain, RAG, tool calling, a ReAct agent with a Gradio chat UI, an MCP server, short- and long-term memory, and a second agent reached over A2A. The lessons come from Scaler Academy's AI Bootcamp, ported to Google Gemini through LangChain. The Blog link opens the write-up of every command, error and fix from my run.
·32 min read·Data Engineering
Data Engineering SQL · Part 6
Part 6 of a data engineering SQL practice bank: six end-to-end capstones, from loading a warehouse in dependency order and a universal SCD2 loader to a late-data-safe incremental pipeline, point-in-time history, a metadata-driven quality gate and a growth mart at the right grain. Every answer is run on real sample data.
·43 min read·Data Engineering
Data Engineering SQL · Part 5
Part 5 of a data engineering SQL practice bank: stand-alone problems from healthcare, IoT, ride-share, telecom, ad tech, logistics, energy, markets and master data, covering interval merging, sweep lines, tiered rating, range and as-of joins, interpolation and survivorship. Every answer is run on real sample data.
·66 min read·Data Engineering
Data Engineering SQL · Part 2
Part 2 of a data engineering SQL practice bank: a retail star schema built question by question, with grain, SCD Types 1 to 3, point-in-time joins, race-free watermarks, partition reloads, inferred members, bridge tables, snapshots and quality gates. Every answer is run on real sample data.