May 2022 – May 2025 · 3 yrs · Bengaluru, Karnataka, India

• Led the design and implementation of Non-White Background (NWB) Project for Fashion products on
• Amazon.in, which was further scaled globally as a multi-tenant service. The project encompasses:
• ∗ Offline Pre-computation Flow: Automated end-to-end pipeline for product ingestion, multi-model ML-based image generation, manual audits, and publishing to media hosting servers. Integrated ETL, Step Functions, Sagemaker, Glue, and Lambda - processing up to 2.5M products globally per week at a cost of $0.002 per product.
• ∗ Online Retrieval Flow: Enabled seamless real-time retrieval of NWB Images metadata with a per region and per tenant peak TPS of 4.8K and a p99 latency of 33ms - leveraging ECS, Caches and DynamoDB.
• It is projected to generate $150M+ yearly global revenue, with a significant boost to search and detail page CTR.
• Co-led Premium Brand-based personalization for Amazon.in Fashion Search by developing an online feature store utilising Redis ElastiCache and DynamoDB, achieving p99 latency of 20ms and peak TPS of 6.2K. The offline pre-computation layer uses ML models and ETL workflows to derive customer-brand affinities. At query time, these affinities are injected into the Lucene Partition response by the retrieval layer, enabling Search rankers to dynamically re-rank products based on the customer’s preferred brands, resulting in a projected $4.2M yearly revenue in India.
• Was last leading design to enhance Ranking and Relevance for Amazon.in Fashion, integrating
• multi-objective ranking using the feature store. The system would host product and customer signals like Sales Velocity, Review Sentiment, Trending and Seasonal Relevance, etc. to dynamically rank results on Search and Detail Pages - driving personalized product discovery, with projected boosts in conversion rates and revenue.
• Drove an org-level campaign to adopt Step Functions Distributed Map for handling large batch workloads and DynamoDB single table design to improve throughput and latency

Aug 2021 – May 2022 · 9 mos

• Architected and shipped high-performance Distributed Storage Volume APIs (1.5K TPS) with enterprise-grade features including Idempotency, RBAC, Concurrency control, and Observability. Leveraged Spring Cloud to implement versioned REST APIs with circuit breaking and audit logging, achieving p99 latency of 320ms.
• Co-led the development of Castor - a single-cluster Control Plane for Distributed Storage Volumes, implementing asynchronous RPCs for critical use cases such as file systems, databases, and disaster recovery.
• Co-led development of Pollux - a multi-cluster Control Plane for Distributed Storage Volumes, implementing backward-compatible fanout RPCs and policy-based automation across the storage fleet.
• Contributed to development of a new API Dev platform (meant for auto-generation of versioned API interfaces and SDKs for multiple teams via plugins built on top of Swagger YAMLs), while ensuring backward compatibility across versions.
• Have helped with interviews for Campus and Team Hiring.

Aug 2020 – Jul 2021 · 11 mos

• Managed Volume Group (equivalent to AWS EBS) Control Plane
• Worked on introducing Volume Group APIs using the versioned V4 Stack on Prism Central.
• Helped teams with onboarding the V4 APIs on the all-new Spring Cloud and Spring Boot based ADONIS API Gateway.
• Catering to the AFS (File Service), DR (Disaster Recovery), and Era (DB Offering on top of Nutanix cluster) requirements w.r.t Volume Groups.
• Heavily involved with Storage Management on PC, which will introduce VG and Storage Container operations on the PC UI.
• Was core reviewer and setter for the Campus Hiring.

Dec 2018 – Jul 2020 · 1 yr 7 mos

• Worked on an entirely new family of Versioned REST APIs for Volume Group (equivalent to AWS EBS) Service which are Namespace and Version Specific. The Custom Maven Plugin requires the developers to just write the annotated YAMLs, and it converts those into DTOs/Interfaces and Client SDKs in Go and JAVA.
• Implementing vDisk container reassignment on AHV for Volume Groups.
• Worked on service to disallow concurrent External Attachments and Load Balance the VM Attachments for the VGs.
• Worked on validation of the Container before its deletion by implementing a check w.r.t presence of VGs.
• Was core reviewer and setter for Campus Hiring

May 2018 – Dec 2018 · 7 mos · Bengaluru Area, India

• Worked on upgrading the client onboarding process along with contract pre-matching using REST based web operations, Spring, Hibernate, Dropwizard and React/Redux.

Jan 2018 – Apr 2018 · 3 mos · Bengaluru Area, India

• Worked on upgrading the creation of accounts and mirror accounts by shifting it from messaging systems to Firm's proprietary workflow management service knows as E-Task, using REST based web operations and ReactJS.

May 2017 – Jul 2017 · 2 mos · Gurgaon, India

• UAID Generation for the Campus Stores, for bulk tagging sort center deliveries for the Campus Program.
• Access Point Report Generation using Admiral

Jan 2017 – Dec 2017 · 11 mos · Allahabad Area, India

• Teaching Assistant for the 1st and 2nd year juniors for courses of Data Structures and Algorithms to prepare them for ACM ICPC.

Jun 2016 – Aug 2016 · 2 mos · Bengaluru Area, India

• Worked on setting, testing, curation and writing editorials of algorithmic contests held on DoSelect platform.

Mar 2016 – Apr 2017 · 1 yr 1 mo · Allahabad Area, India

• Conducted workshops, basically related to competitive coding for the practice of college students, and organized several intra-college coding contests on Hackerearth and Codebuddy. Worked on setting and organizing HUMBLEFOOL CUP 2016, ALKHWARIZM 2017 and CODERED 2018- the flagship events of IIIT Allahabad’s Technical Festival – APAROKSHA. All the contests witnessed great participation from all over India.

Jul 2015 – Mar 2016 · 8 mos · Allahabad Area, India

• Under the supervision of the coordinators, carried out conduction of workshops and coding contests on Hackerearth for practice.