D Words by Dinesh
D

Hi, I'm Dinesh.

Sai Dinesh Reddy Maluchuru · Hyderabad, India

I'm a Data & ML Engineer at Microsoft, working on OneNote Copilot and the OneDrive/SharePoint platform. This is my personal blog — the place where I think out loud about the systems I build and the ones I admire.

What I work on

By day, I help run telemetry and evaluation pipelines that touch 200+ TB of logs and 300B+ transactions every day. I built an LLM-as-a-Judge evaluation framework for OneNote Copilot's video features, migrated several Oracle workloads to Spark, and have spent more time than I'd like to admit chasing 70% cost reductions on petabyte-scale jobs.

Before Microsoft I was a Data Engineer at Amazon (2019–2022), and before that an Associate at JPMorgan Chase (2017–2019). I did my Dual Degree (B.Tech + M.Tech) in Electrical Engineering at IIT Madras from 2012 to 2017.

Why this blog exists

Most of what I read about data and AI engineering online is either thin marketing or dense academic. I wanted a place to write the things I wish someone had handed me earlier: the messy middle — what actually breaks at 3 AM, why a "simple" RAG system needs six iterations before it stops hallucinating, what a 70% cost cut on a query engine really involves.

I publish slowly. Usually one or two essays a month, sometimes none. I try to write only when I have something I'd want to read myself.

What you'll find here

  • Data engineering at scale — Spark internals, lakehouse design, query engines, the unglamorous plumbing
  • AI & LLM systems — RAG, evaluation, prompt engineering, what works in production vs. demos
  • Distributed systems — consensus, storage, networking, the parts most posts skip
  • Honest reflections — bills, postmortems, second-system effects, things I got wrong

Outside work

I got married to Anusha on June 8, 2025 — we're now planning a small AI / robotics startup together (more on that some day, on this blog). I read more than I write, mostly long essays and the occasional textbook. Hyderabad is home.

Stay in touch

The best way to follow new essays is the free newsletter — one email per post, no spam, unsubscribe with one click. There's also an RSS feed if you prefer that. For anything else, the contact page has my email.

This blog is hand-built with Flask + Tailwind. No trackers, no ads, no popups — just words.

Subscribe

One email per essay. No newsletter empire, no roundups, no marketing.