Skip to content

Hey, I'm

Harsh Gupta

I work on distributed systems

Software Development Engineer II at Amazon Web Services — 5 years of building distributed systems that handle tens of millions of requests. I lead teams, design architectures, and ship things that work at scale. Seattle.

Harsh Gupta

About

I care about two things: building systems that don't break under pressure, and the people I build them with. At AWS, I own services that process tens of millions of requests, ship infrastructure across sovereign cloud regions, and design security systems that protect thousands of accounts. I've taken monoliths apart and put them back together as something better — without dropping a single request along the way.

My path here wasn't linear. I started freelancing in Mumbai, moved to Sacramento for my M.S. in Computer Science (specializing in generative AI), then worked through Intel, a startup in New York, data platforms at Bolt in San Francisco, and finally Amazon in Seattle. Each stop taught me something different about building software that matters.

Outside of work, I'm usually hiking somewhere in the Pacific Northwest, experimenting in the kitchen, at the gym, or building something new just to see if I can.

Experience

Amazon Web Services

Apr 2025Present

Software Development Engineer II

Bellevue, WA

  • Spearheaded cross-team delivery of 8 services for a new sovereign cloud region as technical lead — defined the end-to-end architecture, orchestrated a team of 6 engineers, and drove alignment across 5+ platform teams to deliver all services on schedule for the region's public launch.
  • Architected a custom authorization layer to close a critical security gap where users could bypass access controls via an alternate API path — evaluated 3 architectural approaches, performed blast-radius analysis across 1,800+ affected accounts, and deployed incrementally to production with zero customer impact.
  • Pioneered agentic AI workflows with MCP server integrations for CI/CD automation — enabling touchless deployment, auto-remediation, and reducing manual operational overhead across the team's 8+ service portfolio.
JavaKotlinAWS CDKDynamoDBIAM/STSMulti-RegionCI/CD Automation

Amazon — Buy With Prime

Aug 2022Apr 2025

Software Development Engineer

Seattle, WA

  • Designed and drove the migration from a monolithic architecture to distributed microservices — implemented a CDN-based traffic routing layer as the new service entry point, achieving production-readiness for zero-downtime cutover with improved multi-region availability.
  • Built a version-agnostic GraphQL API client adopted by 4+ downstream services — featuring token-based auth, automatic retry with exponential backoff, and error-translation middleware that reduced integration failures and developer toil across merchant-facing systems.
  • Owned and operated scalable event-driven microservices on ECS Fargate and Lambda — handling high-throughput event processing with end-to-end observability, automated deployments, and fully codified infrastructure (IaC).
TypeScriptGraphQLCloudFrontECS FargateLambdaEvent-DrivenIaC

Bolt

Jan 2022Aug 2022

Software Engineer — Data Platforms

San Francisco, CA

  • Built and optimized production data pipelines powering analytics and ML workflows — leveraging BigQuery, Airflow, and Spark. Contributed to an open-source protobuf-to-JSON-schema compiler plugin adopted externally.
PythonBigQueryAirflowSparkData Pipelines
+ Earlier roles

Full-Stack Software Engineer Intern

Intel Corporation · Aug 2021Dec 2021

Full-Stack Software Engineer Intern

The New School · May 2021Aug 2021

Teaching Assistant

CSU Sacramento · May 2019Jun 2021

Student Data Analyst

California Energy Commission · Jul 2019Aug 2019

Trainee Software Engineer

Clairvoyant LLC · Feb 2018Jul 2018

Backend & ML Engineer

Freelance · Oct 2016Feb 2018

Projects

Things I've built and explored.

Batoonee — GAN-Based Conversational AI
Project 01

Batoonee — GAN-Based Conversational AI

My master's thesis. I trained a Wasserstein GAN to generate conversational responses — pitting a generator against a discriminator until the model could answer questions from context. Best architecture hit BLEU 59.7 on Stanford CoQA. The kind of project that teaches you patience.

PythonTensorFlowWGAN-GPSeq2SeqNLP
Emotion Recognizer — Action Unit CNN
Project 02

Emotion Recognizer — Action Unit CNN

Can a machine read your face? I built a chained classifier that watches muscle movements frame-by-frame, predicts which action units fired, then maps that to an emotion. Multi-class, multi-label, surprisingly accurate.

PythonKerasCNNComputer Vision
Guardian — Women Safety App
Project 03

Guardian — Women Safety App

Born from anger after the 2012 Delhi incident. An Android app that sends SOS to the nearest first responder via GPS and alerts your family simultaneously. My bachelor's project — and the reason I became an engineer. Published as a research paper.

AndroidJavaGPSFirebase