Connected context
Integrated authorized documents, tickets, repository tools and saved project work into agent workflows. Made retrieval status and source references visible so missing access is not mistaken for missing information.
Austin, Texas
I build reliable enterprise integrations and AI-assisted engineering workflows.
I work where data, business systems and engineering decisions meet. My focus is making those connections useful, resilient and understandable.
Selected work
Reliable Event Filtering for Enterprise Integrations
An original, synthetic demonstration of change detection, conservative filtering and duplicate delivery handling.
The engineering question
Which record updates deserve downstream work? A fictional customer pipeline receives every update, including housekeeping changes. The policy must reduce unnecessary work without silently discarding a real business change.
The decision
Suppress only changes we can confidently classify as irrelevant. Publish business changes, preserve mixed updates, and route uncertainty visibly.
Built with fictional records. This is not a reconstruction of an employer incident or a claim of production savings.
Professional engineering experience
From an AI answer to a reviewable engineering workflow.
I built and iterated on internal AI-assisted tooling that connects project context, engineering artifacts and test execution. My contribution spans integrations, execution and reporting paths, troubleshooting and the user experience around them.
Implementation stack
Backend and agents: Python, FastAPI, Google ADK and Google Gemini.
Interface: TypeScript, React and Next.js, with CopilotKit / AG-UI for agent interaction.
Delivery: Docker containers on Google Cloud Run, with GitHub Actions for CI/CD.
Deployment stage
Deployed for internal production use, with ongoing feature validation and operational iteration. My work includes implementation, release troubleshooting and post-deployment checks. Individual integrations depend on configured permissions and execution targets.
The problem
Generating an answer is only one step. Engineers still need the right context, a way to review proposed work, a configured execution target and evidence of what actually happened.
QA execution and test visibility
Built workflows for test generation and reuse, configured repository execution and bounded connector assertions. Brought run status, failures and result publishing into the engineering conversation.
Review and integration controls
Improved artifact review, publishing paths and connector setup. Kept interactive guidance connected to real system permissions and environment checks.
Agent guidance and operational visibility
Added reviewed workspace guidance, rules-based model selection and usage/cost visibility. Kept applied guidance distinct from model training, and estimates distinct from billing records.
Understand the requestReview the proposed workRun against a configured targetInspect and share evidence
The engineering judgment
A tool being available does not prove that its credentials work. A generated test does not prove that it ran. A success message does not replace a run report. I focused on making these boundaries clear and diagnosing failures across the workflow.
How I approached verification
Used code review, focused tests, browser walkthroughs and backend/run-log inspection to investigate integration failures and check the resulting behavior. Internal implementation details and operational metrics are not included in this public account.
This is a general account of my professional contribution. The runnable event-filtering project above is a separate, original public demonstration.
Data and integration engineer with approximately three years of professional experience across enterprise integrations, cloud pipeline reliability and applied AI tooling. Experience investigating cross-system failures, implementing resilience improvements, validating data consistency and building AI-assisted engineering workflows.
At Commerce, my official title is AWS Cloud Integration Engineer, with senior data engineering responsibilities.
- Enterprise integrationsInvestigating cross-system failures, defining meaningful change and collaborating on UAT validation.
- Cloud pipeline reliabilityDiagnosing dependency failures and implementing bounded retries and timeouts.
- Data consistencyPreparing reconciliation outputs and QA handoffs across connected systems.
- Applied AI toolingBuilding AI-assisted engineering workflows with connected context, sandbox execution and visible results.
Problems worth investigating
The next questions.
Research directions informed by integration work. These are business hypotheses to test, not validated customer demand or launched products.
CRM-to-billing reconciliation
Help operations teams distinguish a genuine mismatch from timing, rounding and incomplete data.
First validation: ask how teams investigate discrepancies today, what evidence they need and what they would pay to change.
Integration maintenance
Make credential expiry, schema drift and retry behavior visible before an integration becomes a support incident.
First validation: map recent failures, ownership gaps and the limitations of existing monitoring.
AI cost and quality assessment
Compare model choices against a fixed task set, measuring output quality alongside cost and latency.
First validation: define a decision teams struggle to make and the evaluation evidence they trust.