Headline

SDE L5 at Amazon | Backend Engineering | VIT 2019 | Java and Python | DSA | System Design | Fullstack Developer

About

• Building Amazon's cross-border architecture that lets customers shop across borders on Amazon: Global Store arcs, Bazaar and Haul apps, and external partner integrations like JD.com, Kettle, and 11st. • Interested in building large-scale distributed systems, algorithmic problem-solving, and lately, making production systems consumable by AI agents (MCP, agentic workflows). • Experienced across backend, data infrastructure, and cloud: Java, Python, TypeScript, Go, JavaScript, Ruby on Rails, SQL, React. • Comfortable with the AWS toolbox: CDK, Lambda, Kinesis, DynamoDB, S3, SNS, IAM, CloudWatch. • Also worked plenty with Kafka, Redis, Elasticsearch, PostgreSQL, MongoDB, Docker, and Git along the way. • Love sports, and TV shows that have good storylines. • Love exploring new places and meeting new people. • Can't live without friends and family. • I am a most-active-when-it's-2:30-in-the-night person. • Want to be the hero who saves the world.

Location

Bengaluru

AI stack

anthropic apis

claude

hugging face

kiro

Amazon Q Developer

JetBrains AI Assistant

Work experience

SDE L5 at Amazon, Aug 2025 - Present · 1 yr 3 mos, Bangalore Urban, Karnataka, India, CrossBorder | Platform Infrastructure, Global Store Core, Search • Lead technical direction for a 15-engineer team across 4 CrossBorder workstreams (platform infrastructure and gateway, Global Store Core, Search, and Security/Compliance/Privacy), owning roadmap, engineering guidelines, growth, delivery, and operations end to end. • Led the Amazon-JD.com integration enabling all US and JP marketplace use cases. Designed and built the multi-marketplace platform layer (credential routing, per-marketplace secrets, webhook source identification, event orchestration) and drove adoption across 24 downstream teams through launch campaign. The integration now handles $2.1M+ in weekly order sales, about $110M a year. • Author or peer-review roughly one system design a week and mentor engineers across the team. • Cut p90 checkout latency of Global Store orders by ~78% by parallelizing line-item processing that previously ran sequentially, without breaking correctness of the shared order document model. • Shipped the Repay feature for cross-border payments so customers can retry failed payments, this eliminated 60% of order failures caused by payment declines. • Wrote reusable AWS CDK constructs (KinesisFirehoseToS3Construct, S3ReplicationSourceConstruct) for general-purpose cross-partition data replication, adopted by other teams with minimal configuration. Everything ships through automated CI/CD with metrics, alarms, and dashboards. • Built MCP servers over Global Store ordering services and integrated the gateway with production ordering APIs, making core ordering systems directly consumable by AI agents on internal AI-native platforms. Tech Stack: Java, TypeScript, AWS CDK, Lambda, ECS, Kinesis Data Firehose, S3, DynamoDB Streams, SQS, SNS, Secrets Manager, IAM, MCP, Spark-based ETL (Cradle), SQL, CI/CD

Senior Software Engineer at Zeta Global, Mar 2024 - Jul 2024 · 4 mo, India, Audience Explorer | Estimator Service | Segmentation • Owned and re-architected the Estimator Service, a core backend system that returns real-time audience size estimates for complex, nested segmentation filters over 2 Billion user profiles. • Designed a hybrid two-tier estimator: a fast path on bitmap indexes and HLL approximate cardinality, with a fallback slow path using progressive SQL planning for long-tail filters. • p99 latency went from 11s to ~700ms and p50 from 2.5s to ~300ms. The demo helped win a key enterprise client. • Led cross-functional work with Product, Data Science, and DevOps to define estimation SLAs, tune Snowflake and PostgreSQL performance, and keep accuracy tolerances aligned with the product. • Authored the system architecture documentation for Audience Explorer services covering service interactions, data flows, and failure points. Onboarding and debugging got 50% faster. • Cut User Exploration query latency by 60% with Snowflake materialized views, batch precomputation in Python, and better PostgreSQL indexing, enabling dynamic real-time analysis. Tech Stack: Java, Python, Ruby on Rails, JavaScript, Kafka, Redis, AWS, PostgreSQL, Snowflake, React, Unix

SDE II at Tata 1mg, Aug 2023 - Mar 2024 · 8 mos, Gurugram, Haryana, India, Search Service | Latency, Caching, and Relevance • Worked with the Merchandising team on the product search service used by millions of users across web and mobile. • Rewrote Elasticsearch queries with inverted indexing and removed redundant aggregations: p99 dropped 60% (~2200ms to ~880ms) and p50 45% (~800ms to ~450ms). • Added Redis caching for frequently accessed queries; cache hit rate rose 35% and Elasticsearch query volume fell 40%. • Fed user context (past purchases, clickstream behavior) into ranking with Elasticsearch boosting and custom scoring. Search-to-purchase conversion went up 17%. • Search query failures dropped 71%, holding reliability under peak traffic. Tech Stack: Python, Java, Ruby on Rails, JavaScript, Redis, Kafka, Elasticsearch, PostgreSQL, MongoDB, Sanic, React, Unix

Software Engineer at CleverTap, Oct 2021 - Nov 2022 · 1 yr 2 mos, Mumbai, Maharashtra, India, Partner Integrations | API Security | Large-Scale Data Export • Built and scaled the partner export system handling terabyte-scale data exports to AWS S3, Amplitude, Segment, Mixpanel, mParticle, and others. • Redesigned the export pipeline around chunk-based processing with MongoDB cursor pagination, Kafka job queues, and Redis-backed caching, so failed chunks retry independently instead of failing the whole export. • Parallelized exports with Java thread pools and distributed worker pools: export time fell from ~5 hours to ~30 minutes at 10x throughput on the same infrastructure. • Owned and delivered multiple partner integration features: Multiple Passcode (header-based authentication for all Consumption APIs), the Mixpanel Import Deletion Layer (recurring and one-off cohort cleanups), and the Segment Destination Integration (CleverTap as a supported destination on Twilio Segment). Tech Stack: Java, Python, JavaScript, Kafka, Redis, AWS, MongoDB, PostgreSQL, Apache Struts, Vue.js

Developer Associate at SAP Labs India, Jan 2020 - Sep 2021 · 1 yr 9 mos, Bangalore, SAP SuccessFactors | Enterprise Benefits Management • Worked in the Benefits team (Globalization Services) on core modules: Insurance, Pension, Savings Plans, Allowances, and Reimbursements for global enterprise clients. • Built a metadata-driven benefits configuration system on the SAP MetaData Framework (MDF) that lets HR teams define country-specific rules without code changes, cutting custom development effort by 70%. • Led design and rollout of a UK-specific pension scheme: a decorator/wrapper architecture injected region-specific logic into legacy systems without touching core pension flows, with feature-flagged tenant rollout and full backward compatibility. Other geographies later reused the pattern. • Delivered UI enhancements in SAP UI5, optimized OData REST APIs, and added lazy loading; benefits selection page load times dropped 30%. Tech Stack: Java, Python, JavaScript, Spring, MDF, SAP UI5, SAP Cloud, Redis, Jenkins, Unix

Senior Tech Associate at Bank of America, Jun 2019 - Jan 2020 · 7 mo, Chennai

Education

Bachelor of Technology, minor in Computer Science, Vellore Institute of Technology

Software Development, Problem Solving and System Design, Scaler

Public links

linkedin.com: https://www.linkedin.com/in/shivendrasinghprofile