Open source · AWS
Bringing Amazon Bedrock AgentCore to Apache Airflow
How native Apache Airflow operators bring the lifecycle of Amazon Bedrock AgentCore runtimes into DAGs, from creation and invocation through updates and deletion.
Writing
Writing about Data & AI systems.
Practical essays about Data & AI architecture: what we built, why the system took that shape, where the boundaries belong and what a useful proof of concept should actually prove.
5 essays
Architecture, with the decisions left in.
Agents · Airflow
Agentic Airflow: Using AI Agents to Troubleshoot DAG Failures
A practical workflow where Airflow sends failure context to an AI agent that reads the DAG, proposes a fix, opens a draft pull request and notifies the team.
'%3e%3cstop%20offset='0'%20stop-color='%235A9FD4'/%3e%3cstop%20offset='1'%20stop-color='%23306998'/%3e%3c/linearGradient%3e%3clinearGradient%20id='python-original-b'%20gradientUnits='userSpaceOnUse'%20x1='209.474'%20y1='1098.811'%20x2='173.62'%20y2='1149.537'%20gradientTransform='matrix(.563%200%200%20-.568%20-29.215%20707.817)'%3e%3cstop%20offset='0'%20stop-color='%23FFD43B'/%3e%3cstop%20offset='1'%20stop-color='%23FFE873'/%3e%3c/linearGradient%3e%3cpath%20fill='url(%23python-original-a)'%20d='M63.391%201.988c-4.222.02-8.252.379-11.8%201.007-10.45%201.846-12.346%205.71-12.346%2012.837v9.411h24.693v3.137H29.977c-7.176%200-13.46%204.313-15.426%2012.521-2.268%209.405-2.368%2015.275%200%2025.096%201.755%207.311%205.947%2012.519%2013.124%2012.519h8.491V67.234c0-8.151%207.051-15.34%2015.426-15.34h24.665c6.866%200%2012.346-5.654%2012.346-12.548V15.833c0-6.693-5.646-11.72-12.346-12.837-4.244-.706-8.645-1.027-12.866-1.008zM50.037%209.557c2.55%200%204.634%202.117%204.634%204.721%200%202.593-2.083%204.69-4.634%204.69-2.56%200-4.633-2.097-4.633-4.69-.001-2.604%202.073-4.721%204.633-4.721z'%20transform='translate(0%2010.26)'/%3e%3cpath%20fill='url(%23python-original-b)'%20d='M91.682%2028.38v10.966c0%208.5-7.208%2015.655-15.426%2015.655H51.591c-6.756%200-12.346%205.783-12.346%2012.549v23.515c0%206.691%205.818%2010.628%2012.346%2012.547%207.816%202.297%2015.312%202.713%2024.665%200%206.216-1.801%2012.346-5.423%2012.346-12.547v-9.412H63.938v-3.138h37.012c7.176%200%209.852-5.005%2012.348-12.519%202.578-7.735%202.467-15.174%200-25.096-1.774-7.145-5.161-12.521-12.348-12.521h-9.268zM77.809%2087.927c2.561%200%204.634%202.097%204.634%204.692%200%202.602-2.074%204.719-4.634%204.719-2.55%200-4.633-2.117-4.633-4.719%200-2.595%202.083-4.692%204.633-4.692z'%20transform='translate(0%2010.26)'/%3e%3cradialGradient%20id='python-original-c'%20cx='1825.678'%20cy='444.45'%20r='26.743'%20gradientTransform='matrix(0%20-.24%20-1.055%200%20532.979%20557.576)'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20offset='0'%20stop-color='%23B8B8B8'%20stop-opacity='.498'/%3e%3cstop%20offset='1'%20stop-color='%237F7F7F'%20stop-opacity='0'/%3e%3c/radialGradient%3e%3cpath%20opacity='.444'%20fill='url(%23python-original-c)'%20d='M97.309%20119.597c0%203.543-14.816%206.416-33.091%206.416-18.276%200-33.092-2.873-33.092-6.416%200-3.544%2014.815-6.417%2033.092-6.417%2018.275%200%2033.091%202.872%2033.091%206.417z'/%3e%3c/svg%3e){kind=small}
Architecture · AWS
Multi-Agents with Bedrock and Claude: One Supervisor and Three Specialists
A routed multi-agent architecture for a learning product, with one supervisor and specialized agents selected for distinct cognitive jobs.
Data platform · AI
Designing a Modern Data & AI Platform on AWS
An end-to-end reference architecture that connects ingestion, transformation, governance, observability and AI-powered analytics.
RAG · Bedrock
POC: Amazon Bedrock + S3 Vectors
A focused proof of concept for grounding an Amazon Bedrock agent on documents through S3 Vectors.