Tejal Shanbhag

About Tejal Shanbhag

I’m a Software Development Engineer with 3.5+ years of experience building distributed systems, cloud-native services, and machine learning–powered applications across E-commerce, Retail Tech, and Conversational AI. My work sits at the intersection of Search, Ranking, and Personalization systems deployed at production scale.At Amazon, I’ve developed Prime acquisition systems and improved Search relevance for emerging marketplaces (India, Mexico, South America) by building Learning-to-Rank (LTR) and re-ranking pipelines that incorporate CTR, dwell time, and conversion features. I evaluate model performance using offline metrics such as NDCG and precision/recall, and validate impact through large-scale A/B experimentation, consistently improving engagement while maintaining strict P99 latency SLAs.I’ve also designed multilingual query interpretation pipelines leveraging Large Language Models (LLMs), embeddings, and Retrieval-Augmented Generation (RAG) architectures to improve recall and intent detection for blended-language queries. On the Alexa team, I built synthetic data pipelines and dialog systems that accelerated NLP model development and improved conversational quality across devices like FireTV.Technically, I specialize in Python, Java, and AWS (Lambda, DynamoDB, EC2, S3, CDK, Step Functions), building scalable distributed ML systems with SageMaker and Amazon Bedrock. My experience includes feature engineering, semantic retrieval, vector-based search, data pipelines (Glue, Athena), and experimentation frameworks, along with observability and monitoring to ensure production reliability.I enjoy building intelligent systems that balance model quality, scalability, and measurable business impact.

Experience

1. Software Engineer

   Amazon

   Sep 2024 — Present · Seattle, WA, US

   Launched localized search ranking features for India, Mexico, Canada, and South America by building serverless pipelines with AWS Lambda, Step Functions, DynamoDB, and S3, orchestrated via AWS CDK, improving click-through rates by 15–20% across emerging markets- Improved search relevance and ranking quality by integrating CTR, dwell time, and conversion signals into ML ranking and re-ranking pipelines using SageMaker, AWS Glue Studio, and Athena for large-scale feature processing and analysis, achieving a 12% lift in relevance metrics while maintaining P99 latency SLAs through controlled A/B experiments- Designed and deployed a multilingual query interpretation pipeline leveraging LLMs on Amazon Bedrock (Claude) for intent detection, normalization, and translation of blended-language queries (e.g, Hinglish), with data persisted in S3 and workflows coordinated via Step Functions, improving recall and accessibility while keeping latency-critical ranking paths LLM-free- Led the migration of production services from DUB to FRA and ZAZ as part of the ExitDUB initiative using active-active architectures, infrastructure defined via AWS CDK, and phased traffic shifting with rollback-safe cutovers, ensuring zero downtime and regional resiliency- Implemented DynamoDB Global Tables for multi-region replication, backed by Lambda-based health checks, automated alarms, and analytics via Athena and QuickSight dashboards, ensuring data consistency, failover readiness, and operational visibility across regions- Owned and enhanced the Prime Acquisition Upsell widget provider using Java, EC2, Lambda, DynamoDB, and S3, with experimentation metrics surfaced through QuickSight, enabling scalable, personalized offers across geographies and customer segments and driving a 10–15% increase in Prime conversion rates.

Education

• South Indian Education Societys College of Arts Science and Commerce Sion West Mumbai 400 022

  HIGHER SECONDARY CERTIFICATE (HSC)

  2011 — 2013

• AUXILIUM CONVENT HIGH SCHOOL

  SECONDARY SCHOOL CERTIFICATE (SSC)

  2001 — 2011

• Usha Mittal Institute of Technology for Women

  Bachelor of Technology - BTech, Computer Science

  2013 — 2017

• Rochester Institute of Technology

  Masters Of Science, Computer Science

  2019 — 2021