About

I'm Swati Sisodia, a Software Development Engineer III at Amazon working on GenAI and agentic systems: agents that generate their own UI, RAG-based support systems built on Amazon Bedrock, and the retrieval and infrastructure that supports them. I've spent my career across the full stack: backend services handling millions of requests per second, GraphQL APIs, multimodal search, and now agentic AI architecture.

Swati Sisodia

Experience

Amazon

July 2017 to Present

Software Development Engineer III

September 2023 to Present

  • Architected an agentic UI generation system using Amazon Bedrock Agents to dynamically generate stateful, bidirectional UI components from LLM outputs, moving beyond static text responses to interactive, context-aware interfaces served via AWS Lambda.
  • Led end-to-end design and build of a RAG-based GenAI support agent that replaced legacy rule-based chatbots, using Amazon Bedrock Agents for orchestration and building the full knowledge base ingestion pipeline (chunking, embedding, and indexing into Amazon OpenSearch). Integrated with ServiceNow and provisioned via AWS CDK, the system automated resolution of 120k annual tickets and reduced human-led ticket volume to 5%.
  • Led design and implementation of an Organic Content Server serving millions of requests; pivoted the architecture to GraphQL to unify field resolution across 5+ owning services, replacing a custom per-placement resolution approach and saving 4+ weeks of engineering effort, with the pattern adopted as the team standard for similar aggregation problems.
  • Architected and led the Promoted Posts initiative, reduced the original 80-week projected timeline to 35 weeks by proposing infrastructure reuse and repositioning product's original scope, then led a team of 7 engineers through launch; scaled to 150,000+ advertisers, generating $4M in revenue.
  • Architected a large-scale multimodal search system using CLIP embeddings for 80M+ product, advertiser, and brand images and their captions, indexed in Amazon OpenSearch and keyed by advertiser/brand ID, enabling advertisers to search and match the right image to their caption when creating new posts.

Software Development Engineer II

August 2019 to September 2023

  • Designed and implemented a high-scale service for Amazon Posts, serving multi-million requests per second across search pages and feeds, while developing desktop/mobile widgets to drive brand discovery. Proposed and led the shift from a legacy monolith to a DynamoDB Accelerator (DAX) microservice architecture after evaluating an existing Redis cluster and validating the tradeoff with data before committing engineering time.
  • Built an automated deployment pipeline for Applied Science models on AWS EMR: a daily cron-triggered service Dockerized and deployed models, spun up EMR clusters to run them, and tracked job execution for success or termination to determine retries, reducing productionization effort by three weeks.
  • Built a metrics pipeline capturing frontend clicks/impressions and server-side impressions via Kinesis, with hourly Glue jobs to clean, aggregate (post-level views, time spent, etc.), and prepare the data for downstream use, powering both business-facing monitoring dashboards and the ML recommendation pipeline.

Software Development Engineer I

July 2017 to August 2019

  • Led a critical migration off Amazon's legacy Perl-based system (Gurupa) to a modern Java architecture, and implemented automated testing for event-driven pipelines, eliminating manual validation, improving system reliability and reducing technical debt.
  • Redesigned the workflow for a self-service demand modeling application used by hundreds of product managers at Amazon, moving scenario execution and custom report generation onto the web app to replace manual Excel-based post-processing, cutting turnaround from about a month to two days.
  • Built Kindle search-page features, including displaying ebook prices inline with search results and enabling customers to subscribe to an author directly from search.

Education

University of Florida

August 2015 to August 2017

Master of Science, Computer Science (GPA 3.5/4.0)

Graduate coursework: Analysis of Algorithms, Advanced Data Structures, Database Management Systems, Social Network Computing, Computer Networks, Distributed Operating Systems, Machine Learning, Introduction to Data Science, Statistics.

Maulana Azad National Institute of Technology

August 2009 to May 2013

Bachelor of Technology (CGPA 8.08/10)

Skills

AI Systems & Agents

LLM Application DevelopmentAgentic Workflows & Tool-callingMulti-step Agent Orchestration (Bedrock Agents)Conversational AI DesignPrompt Engineering & VersioningStructured Output / Function CallingGuardrails & Fallback Design

Retrieval & Knowledge Systems

RAG ArchitectureVector Search & EmbeddingsAmazon OpenSearchHybrid/Semantic SearchKnowledge Base DesignChunking & Indexing Strategies

AWS & Infrastructure-as-Code

AWS CDKAmazon BedrockLambdaECSStep FunctionsAWS GlueEMRSparkDistributed SystemsRedisDockerGraphQLReal-time Monitoring & Metrics Infrastructure

Data Pipelines

Glue-based ETLEvent-driven PipelinesAutomated Testing & Validation

Languages

PythonScalaJavaJavaScriptTypeScriptC++

Get in touch

Contact me onLinkedInor atswati.sisodia61@gmail.com.