My expertise spans

Generative AI Engineer with 6.8 years of professional experience. I lead end-to-end delivery of agentic AI systems for fintech, from architecture and client requirements to production, using Amazon Bedrock, AgentCore and RAG, on top of a strong backend foundation in scalable, reliable financial platforms.

6.8Years of experience
3Featured GenAI projects
20+AWS services used
End-to-endArchitecture to production
  • Generative AI: Amazon Bedrock · AgentCore · LangGraph · Strands SDK · LLMs · RAG · Multi-Agent Systems · Agentic AI Orchestration · SageMaker AI
  • Cloud & AWS: ECS Fargate · S3 · DynamoDB · Polly · Kendra · API Gateway · CloudFormation · CloudTrail · Glue · Kinesis · Secrets Manager · EMR · Transcribe · Athena · QuickSight · RDS · OpenSearch
  • Deployment: Independently architected and deployed production-ready Generative AI projects on AWS, delivering secure, scalable solutions with end-to-end data pipelines from ingestion to visualization.

What I do

Generative AI & Agents

Building conversational and agentic systems that handle real customer workflows: multi-agent orchestration, RAG pipelines, semantic search and prompt orchestration for accurate, context-aware answers.

Bedrock · AgentCore · LangGraph · Strands SDK · RAG

Cloud-Native Backends

Serverless, secure APIs on AWS with API Gateway, Lambda and DynamoDB, protected by Cognito and JWT auth, and observable end-to-end with CloudWatch.

API Gateway · Lambda · DynamoDB · Cognito · CloudWatch

Data Pipelines & Analytics

Event-driven pipelines from ingestion to insight: audio processing with Transcribe, automated cataloguing with Glue and SQL analytics with Athena.

S3 · Glue · Athena · Kinesis · EMR · QuickSight

Knowledge

Generative AI

Amazon BedrockAgentCoreLangGraphStrands SDKLLMsRAGRAG OrchestrationMulti-Agent SystemsAgentic AI OrchestrationSageMaker AILangChain

Cloud & AWS

ECS FargateS3DynamoDBAPI GatewayLambdaCognitoCloudFormationCloudTrailCloudWatchSecrets ManagerGlueKinesisEMRAthenaQuickSightTranscribePollyKendraRDSOpenSearch

Experience

Dec 2022 – Present

Cloud Age Global IT Services LLCGenerative AI Engineer

  • Engineered a conversational banking assistant on Amazon Bedrock AgentCore (Runtime, Gateway, Identity, Policy, Memory), built with the Strands Agents SDK and DeepSeek V3.2 on Bedrock to securely process customer banking queries.
  • Governed money movement with a confirmation gate and Cedar policy limits, with tools exposed over MCP through AgentCore Gateway and Cognito JWT identity.
  • Built cloud-native, serverless APIs with API Gateway, Lambda and DynamoDB, secured with Cognito and JWT auth.
  • Implemented end-to-end observability with CloudWatch, enabling scalable, secure, real-time financial assistance.
Cloud Age Global IT Services LLC

AI Cloud Engineer

Project: RAG Chatbot – AI Knowledge Assistant · Client: Nasdaq

  • Built a RAG Knowledge Assistant with LangChain and Amazon Bedrock Agents for enterprise knowledge retrieval.
  • Implemented semantic search, top-k retrieval and prompt orchestration for accurate, context-aware responses.
  • Improved discoverability, response relevance and team efficiency with a scalable GenAI solution.

Project: Transcribe Assistant · Client: Maximus

  • Architected an event-driven audio processing pipeline using S3, Lambda and Amazon Transcribe.
  • Integrated Glue and Athena for automated data cataloguing and SQL-based querying.
  • Delivered an end-to-end audio-to-insight workflow with automated analytics.
Jun 2019 – Dec 2019

Galaxy Transmission Pvt LtdQuality Engineer

  • Managed quality documentation, inspection reports and ERP records, keeping the team audit-ready and in compliance.
  • Analyzed root causes of defects and drove process improvements with cross-functional manufacturing teams, cutting quality issues by 20%.
Dec 2018 – May 2019

Samruddhi Industries Pvt LtdQuality Engineer

  • Executed RCA-CAPA processes and preventive action planning as a certified Internal ISO Auditor.
  • Managed quality compliance and customer complaint resolution; led FMEA to drive zero-defect (ZD) outcomes.
Jul 2015 – Feb 2017

Hi-Tech Plast LtdQuality Engineer

  • Led QA administration, including supplier and 5S audits, and ran the “Can Do” zero-defect program.
  • Delivered 2 CIP and Kaizen projects with production and engineering teams, ensuring on-time, defect-free delivery.

Portfolio

Certifications & Education

Generative AI CertificationCertification
AWS AI Practitioner ChallengeTraining

Education

B.E.
Bachelor of Engineering

Working skills

Generative AIBedrock · AgentCore · LLMs · RAG
Agent FrameworksLangGraph · Strands SDK · LangChain
AWS CloudECS Fargate · Lambda · DynamoDB · S3
Security & AuthCognito · JWT · Secrets Manager
Data & AnalyticsGlue · Athena · Kinesis · QuickSight

Contact

Have an agentic AI or GenAI project in mind? Let’s talk.

Emailsayalisawant2819@gmail.com
Phone+91 97304 98888
LinkedInsayali-sawant
GitHubsayalisawant-dev

Conversational Banking Assistant on Amazon Bedrock AgentCore

Fintech · retail-banking prototype · mock customer data

Overview

A conversational assistant that answers customers from their own account data, with strict controls on anything that moves money. It understands plain language, handles routine requests without staff help, remembers past requests across sessions, and runs on managed AWS services with no servers to maintain.

Use cases

  • Balances & account overview: checking, savings and card balances, card minimum payment and due date.
  • Spending insights: filters transactions by account or category and summarises income, spending and net cash flow.
  • Disputes: spots duplicate charges, reads the charge back and asks before filing.
  • Transfers between own accounts: validates ownership and limits, returns a reference and fresh balances.
Agent: Strands Agents SDK
Model: DeepSeek V3.2 on Amazon Bedrock
Tools: MCP via AgentCore Gateway (6 read-only, 2 state-changing)
Backend: AWS Lambda (Python 3.12)
Identity: Amazon Cognito user groups with JWT claims
Front end: static web page on AWS Amplify
Deploy & quality: AWS CDK (TypeScript), AgentCore Evaluations, CloudWatch

Security by design

  • Prompt rules: own data only, every number comes from a tool.
  • Confirmation gate: money moves only after the customer says “yes” in a following turn.
  • Cedar policy limits by customer tier, enforced at the AgentCore Gateway.
  • Isolation: a microVM per session and memories kept per user.

Prototype results (Oct 2026)

1.00Goal success (AgentCore evaluator)
8 / 8Tool calls correct
37Automated tests

Prototype built with mock customer data on a single recorded evaluation session. It has no real users and has not been assessed for production use with real financial data.

RAG Chatbot – AI Knowledge Assistant

Client: Nasdaq · AI Cloud Engineer, Cloud Age Global IT Services

Overview

An enterprise knowledge assistant that retrieves answers from organisational content using Retrieval-Augmented Generation.

What I built

  • RAG Knowledge Assistant with LangChain and Amazon Bedrock Agents for enterprise knowledge retrieval.
  • Semantic search, top-k retrieval and prompt orchestration for accurate, context-aware responses.
  • A scalable GenAI solution that improved discoverability, response relevance and team efficiency.
Stack: LangChain · Amazon Bedrock Agents · semantic search · top-k retrieval · prompt orchestration

Transcribe Assistant

Client: Maximus · AI Cloud Engineer, Cloud Age Global IT Services

Overview

An end-to-end audio-to-insight workflow: audio lands in the cloud, is transcribed automatically, catalogued and made queryable with SQL.

What I built

  • Event-driven audio processing pipeline using S3, Lambda and Amazon Transcribe.
  • Automated data cataloguing with AWS Glue and SQL-based querying with Amazon Athena.
  • Automated analytics delivered across the full audio-to-insight workflow.
Stack: Amazon S3 · AWS Lambda · Amazon Transcribe · AWS Glue · Amazon Athena