Gift-store assistant with Bedrock Converse and a custom Lambda tool loop.
Demo access on request
- Converse API + tool loop for search, policy Q&A, and cart
- Knowledge Base + S3 Vectors for cited policy answers
CDK
Bedrock
Lambda
DynamoDB
Same assistant on Bedrock Agents — managed orchestration and action groups.
Demo access on request
- Bedrock Agent with OpenAPI action group Lambda
- Knowledge Base attached for policy retrieval
Bedrock Agents
CDK
Lambda
DynamoDB
Same assistant on AgentCore Runtime, Gateway MCP tools, and Memory.
Demo access on request
- Runtime Converse loop with Gateway tool dispatch
- Short-term Memory for shopper session context
AgentCore
CDK
Bedrock
DynamoDB
Same assistant on Bedrock Mantle Responses API with server-managed state.
Demo access on request
- OpenAI-compatible Responses API on bedrock-mantle
- Stateful turns via previous_response_id
Mantle
CDK
Bedrock
Lambda
API Gateway + Lambda + DynamoDB, tuned across six memory configs.
- 66× faster and 12× cheaper at 512 MB vs 128 MB
- Two CDK stacks: runtime API and Step Functions tuner
CDK
Lambda
DynamoDB
Step Functions
Python
Ten-phase CDK lab: Cognito, IoT, SQS, analytics, and live insights.
- Live dashboard for steps, activity, and heart rate
- Cognito auth, SQS ingest, CloudWatch, and WAF
CDK
Lambda
Cognito
IoT
SQS
WAF
Compares in-memory, DynamoDB, and Redis when Lambda traffic fails over.
- Three storage modes with simulated mid-session failover
- Live visualizer of session carry-forward vs loss
CDK
Lambda
DynamoDB
API Gateway
CloudFront
Kinesis on-demand → Lambda → DynamoDB with a live activity API.
- On-demand Kinesis with partial batch failure retries
- Atomic product counters for view, cart, and purchase
CDK
Kinesis
Lambda
DynamoDB
API Gateway
Scores Greenhouse/Lever URLs for stale listings with a live paste-a-URL check.
Demo access on request
- 0–100 risk score from age, reposts, and careers cross-check
- Daily collectors + 12h check cache (~$3–10/mo)
CDK
Lambda
DynamoDB
SQS
API Gateway
EventBridge
LangChain/LangGraph + Bedrock turns questions into Athena SQL over S3 lakes.
- Cut ad-hoc reporting from days to minutes
- Self-service analytics for leadership without BI wait
LangChain
LangGraph
AWS Bedrock
Athena
S3
Python
Production RAG with FAISS + Bedrock embeddings on ECS Fargate.
- Semantic search with FAISS and Bedrock embeddings
- IAM, VPC isolation, WAF v2, and Secrets Manager
RAG
FAISS
AWS Bedrock
ECS Fargate
WAF v2
Python
CDK + GitLab CI with STS AssumeRole — no long-lived credentials.
- Instance profile → tagged STS AssumeRole across accounts
- Policy factories, permission boundaries, and cdk-nag
CDK
GitLab CI
IAM
STS
cdk-nag
TypeScript
PySpark pipelines: SQL Server → S3 Parquet + Athena + Glue.
- Nightly full-replace sync via EventBridge Scheduler
- Unlocked self-service BI across business teams
PySpark
Athena
Glue
EventBridge
S3
Python
End-to-end AWS with Terraform workspaces and DynamoDB state locking.
- Zero config drift across dev/test/prod
- ECS, RDS, Cognito, Lambda, WAF v2, Glue, Athena
Terraform
ECS Fargate
RDS
Cognito
WAF v2
DynamoDB
GitLab → Docker/ECR → ECS rolling revisions on every commit.
- Zero-downtime ECS task definition revisions
- Claude Code for agentic workflow prototyping
GitLab CI
Docker
ECR
ECS
IAM
Claude Code