Harsh Gupta

Software Development Engineer II

Bellevue, WA

Moved to AWS to take on larger, multi-team programs with visible customer impact — region launches, public-API architecture, and the kind of operational work that unlocks whole orgs. The through-line is stepping into ambiguous problems and shipping them cleanly.

• Tech lead on an eight-service sovereign-cloud region launch, inherited at month two of three. Built a team of five engineers from scratch, drove cross-org alignment through the principal-engineer and senior-manager layers, and unblocked dependencies across five platform teams. Personally launched the most architecturally complex service in the set. All eight shipped on deadline with validated readiness reviews — available from day one of the region’s public launch.
• End-to-end owner on a multi-year authorization redesign for a public AWS API. Evaluated three architectural approaches, implemented a Lambda-authorizer strategy with partial authorization for overlapping API surface, and drove a Priority-1 product-security engagement through penetration testing to full sign-off. Rolled out in 20% increments with zero customer regression; unblocked three dependent programs.
• Self-selected and closed out a stack of operational risk nobody had owned — migrated long-lived credentials off the legacy auth scheme across 16+ accounts and 7 partitions (all risks resolved before escalation), modernized two CDK pipelines to the latest region-build patterns (one-click multi-partition deploys), and fixed an alarm-replication gap silently hiding on-call signal for months. On-call incident response dropped from hours to minutes.
• Currently building agentic-AI utilities on Bedrock and SageMaker to automate repetitive operational-review and readiness work for the team.

Software Development Engineer

Seattle, WA

Joined as an early engineer on a new org inside Amazon and grew into the SDE who owned the parts of the stack nobody else wanted — CI/CD, API integration layers, production auto-heal. The bet was that making the boring parts work beautifully would compound for the whole team.

• Designed and shipped the team’s dedicated GraphQL client for its third-party platform integration — a version-agnostic Kotlin package with targeted retry strategies, precise status-code handling, and observability built in. Replaced ad-hoc query code scattered across services and eliminated the recurring "API version upgrade" operational load each quarter.
• Led the full CI/CD transformation of two merchant-onboarding UI pipelines. Built a reusable CDK construct (adopted by other teams across the org) that runs integration tests as parallel approval steps — cut average run time from 35 minutes to 11, a 68.5% reduction. Deploy cadence moved from 2-3 per month to 1-2 per week; manual interventions dropped from 15-20 per month to 5-10; eliminated ~3-4 engineering hours per deploy previously spent on change management. Recognized as a team-wide operational win.
• Shipped an auto-heal system inside seven days for a post-launch shipping-configuration bug that was causing incorrect merchant order routing. Delivered the autocorrection logic, integration-test harness, and a real-time dashboard surfacing failure counts and correction rates — closed the failure mode at the source.
• Designed a CloudFront-based unified service ingress as the monolithic backend split into a fleet of domain-specific services — path-based routing, clear separation between application teams and infrastructure ownership, and a migration plan for incremental cutover.

Software Engineer — Data Platforms

San Francisco, CA

First post-grad role in San Francisco. Got a fast apprenticeship in production data systems and shipped against real SLAs — a strong runway into the Amazon work that followed.

• Built and tuned production data pipelines powering analytics and ML workflows on BigQuery, Airflow, and Spark. Contributed to an open-source protobuf-to-JSON-schema compiler plugin.