SDE II at Amazon · Gurugram, India
Software Development Engineer II at Amazon, four years in. My work has spanned Finance Tech services running at millions of transactions per second and Alexa, where I am currently migrating Alexa Daily Essentials onto Alexa+, Amazon's Gen AI assistant. On the side I write and build learn.sauravgpt.in, and ship mobile apps that people actually use.
Two things I own end to end, and two at Amazon where the interesting part is the scale rather than the code I can show you.
Live Built & written solo
A course platform where I write every lesson myself. Three tracks (System Design, Generative AI, and Data Structures & Algorithms), each with per-lesson progress tracking and resume-where-you-left-off.
Scope: 63 lessons across 12 modules, written from production experience rather than textbooks.
Google Play Published
An all-in-one pregnancy companion: week-by-week milestones for mother and baby, a health, mood and medicine log, a kick counter, trimester-aware recipes, and appointment notes kept in one place.
Design call: privacy first, so personal health logs stay on the device rather than on a server, and there is no social feed.
Amazon 2024 to present
Migrating Alexa Daily Essentials experiences onto Alexa+, Amazon's LLM-powered assistant. The work is less about prompts and more about re-modelling deterministic intent handlers as agentic, tool-calling flows without regressing the latency and reliability people expect from a daily-use assistant.
The hard part: keeping a non-deterministic model inside latency and accuracy budgets that were written for deterministic code.
Internal to Amazon, no public repository.
Amazon Finance Tech
Backend services in Amazon's Finance Tech org operating at millions of transactions per second. At that volume idempotency, safe retries and graceful degradation stop being good practice and become the only way the system stays correct.
What it taught me: at this scale every failure mode is a certainty, so you choose which ones you can live with in advance.
Internal to Amazon, no public repository.
No projects match that filter yet.
Most system design material is either a listicle or a 900-page book. learn.sauravgpt.in is my attempt at the middle: short, engineering-first lessons that build in order, written by someone who is still on call. The lesson content is open source, so corrections and new lessons are welcome.
Foundations (client-server, TCP/HTTP, caching, replication), non-functional requirements (availability, latency, consistency, durability), core building blocks (load balancing, CDNs, rate limiting, idempotency, pub/sub), then the data layer (indexing, consistent hashing, CAP, partitioning, sharding, tail latency).
How modern generative AI actually works: LLM internals, prompting, RAG, agents, and what changes when you put any of it into production.
Interview-grade DSA from the ground up: complexity analysis, core data structures, trees and graphs, and the handful of algorithmic patterns behind most coding problems.
Progress is tracked per lesson and the site resumes where you stopped. Built with Next.js so lesson pages stay static and fast, because a learning site that takes three seconds to load does not get read.
Lessons live in a public repo at sauravgpt/sauravgptin-content. Fixing a wrong claim, tightening an explanation or adding a lesson is a pull request away.
Oct 2024 to Present
Amazon · Gurugram, India
Jul 2022 to Sep 2024
Amazon · India
Jan 2022 to Jul 2022
Amazon · Chennai, India
May 2021 to Jun 2021
XPCover · India
Jan 2021 to Mar 2021
MobbyPark · Bengaluru, India
Grouped by what I would be comfortable being questioned on, rather than everything I have ever installed.
Happy to talk about Gen AI in production, system design, or anything at the awkward intersection of scale and correctness.