Seven years of building on AWS.

Architecture and advisory

• Ran technical discovery on business problems, data readiness and AI use cases, then turned them into solution designs, implementation plans, estimates and proofs of concept, and handed over production solutions with deployment guides and runbooks.
• Advised international organizations and global development institutions on modernization, workload migration and production readiness across networking, identity, compute, storage, databases, serverless, observability and governance.
• Designed and reviewed highly available workload patterns with VPC, EC2, Elastic Load Balancing, Auto Scaling, RDS, Lambda, Route 53 and CloudWatch, weighing failure modes, recovery, scaling and operational ownership, with experience across Microsoft Azure services and architectures.
• Built full-stack and containerized solutions with AWS Amplify, API Gateway, AppSync, DynamoDB, Lambda, Docker, Amazon ECS and Amazon ECR.

Generative and agentic AI

• Built RAG assistants on Amazon Bedrock Knowledge Bases with Amazon S3 data sources, chunking and embedding strategies and vector search, measuring quality, accuracy, latency and cost while iterating on prompts, retrieval and model selection.
• Built AWS Lambda integration and orchestration layers that connect Bedrock models and agents to applications, REST APIs, data sources and third-party systems, secured with Amazon Cognito and least-privilege IAM, with human approval for sensitive agent actions.
• Designed Amazon Connect self-service with Amazon Lex bots fulfilled by Lambda and Bedrock, plus AI-assisted agent experiences such as call summaries and knowledge answers grounded in customer content.
• Developed ingestion, embedding and retrieval pipelines with Amazon S3, AWS Glue and Amazon Athena in Python and SQL, and used Amazon SageMaker for training, fine-tuning and endpoints where foundation models alone fell short.
• Established an agent-platform architecture on Amazon Bedrock AgentCore that runs LangGraph, Strands Agents, OpenAI Agents SDK and Google ADK agents under shared security, governance, observability and deployment controls, adopting MCP and A2A where they fit.
• Defined the controls every agent inherits: AgentCore Identity for inbound and outbound authentication, least-privilege execution roles, Gateway-mediated and policy-governed MCP tool access, credential isolation, memory namespaces with retention controls and Amazon Bedrock Guardrails.
• Instrumented agents with AgentCore Observability and OpenTelemetry traces in Amazon CloudWatch, paired with AgentCore Evaluations for correctness, tool selection and safety, so teams could review behavior, cost and latency before and after release.
• Worked hands-on with engineering teams on reference agents, Runtime and Gateway patterns, MCP servers, memory, CI/CD, infrastructure as code and reusable templates that shortened the path from prototype to production.

Enablement

• Designed and delivered a hands-on AgentCore enablement program for the engineering teams of a global development institution, including a live-deployed concierge agent on AgentCore Runtime and Memory that remembers users across sessions and rebuilds with zero code changes.
• Authored AgentCore Runtime and Memory deep-dive sessions and an Amazon Translate technical brief for international organization stakeholders.
• Ran architecture reviews, immersion days, migration-planning sessions and workshops across healthcare, nonprofit, education, law-enforcement and international public-sector teams, explaining AI, security and architecture tradeoffs to executives and engineers.

Security

• Member of the AWS Security Technical Field Community, helping teams harden workloads through IAM and least privilege, MFA, network segmentation, security groups, logging, monitoring, responsible AI and secure infrastructure as code.
• Worked with engineering teams on VPC connectivity, VPN, routing, DNS, security boundaries, monitoring and alerting during architecture reviews and implementation planning.
• Clarified trust boundaries, identified architecture risks and shaped controls that teams could operate after handoff.

Agent platform work, 2025 to 2026

• Created AgentCore foundations with the AgentCore CLI and AWS CDK, using source-controlled configuration for runtimes, memory, credentials, gateways, evaluators, policies and deployment targets.
• Configured a Python and LangGraph agent runtime with CDK-managed infrastructure, and designed memory strategies for semantic facts, preferences and session summaries with namespace separation and retention controls.
• Built a live deploy bridge that streams real AgentCore CLI events onto an interactive architecture diagram, with a serverless chat Lambda, a runtime status API and Terraform-managed Amazon S3 and CloudFront hosting.
• Packaged the patterns as reusable developer templates so every new agent starts from a secure, observable, deployable baseline.