Move single-tenant Search clusters to shared model-serving infrastructure across tenants and locales.
My contribution
Chose strong tenant resource isolation with selected shared capabilities, designed multi-tier inference routing, and rolled out the migration incrementally with the team, validating isolation and failover before expanding traffic.
Deployment outcome
Deployment practices associated with the architecture contributed to a 40% improvement in deployment cycle time.
Beyond the migration
Built query normalization and tokenization adopted across Search teams, and established architecture governance and distributed systems training.
PayPal 2023 to 2024
Financial risk. Faster decisions.
Engineered a GCP reporting platform for fraud detection and chargeback analytics, then modernized the data workflows behind it.
Java
GCP
DynamoDB
Databricks
10M+
daily transactions processed
30%
lower costs for migrated data workflows
Fraud and chargeback analytics for merchant systems.
Explore the engineering
Platform challenge
Support high-volume risk analytics while modernizing legacy data workflows and improving feature delivery.
My contribution
Built the financial risk reporting platform, orchestrated Databricks analytics, and migrated core RDBMS workflows to DynamoDB.
Additional outcomes
Analytics efficiency improved by 30%. ORM-based data-layer modernization reduced the feature-to-production lifecycle by 50%.
Outcome scope
The reporting platform, Databricks orchestration, and data-layer modernization were distinct parts of the work. The cost reduction relates to the RDBMS-to-DynamoDB workflow migration.
IBM 2016 to 2019
Global infrastructure. Operational clarity.
Led a team of 10 engineers running monitoring and alerting for IBM Cloud across 50+ data centers and 10,000 servers.
Kafka
Elasticsearch
Grafana
Java
99.99%
uptime maintained
$3M+
infrastructure cost reductions
A logging pipeline ingesting 500 TB each day.
Explore the engineering
Platform challenge
Maintain visibility and reliable operations across a global cloud infrastructure footprint.
My contribution
Led monitoring with Icinga, Grafana, and PagerDuty. Built a logging pipeline using Logstash, Kafka, Elasticsearch, and Kibana that informed infrastructure cost reductions.
Operational ownership
Led the cross-functional engineering team and served as the subject-matter expert for L1 escalations on a 15-minute SLA.
Outcome scope
The logging pipeline informed the broader infrastructure cost reductions. The $3M+ figure describes that infrastructure outcome, rather than a measured reduction in log-storage cost alone.
Ford / 2019 to 2023
Making cloud infrastructure usable.
I designed the Azure Right-to-Provision service for subscription creation and budget management, built TypeScript role-based access controls and automated cloud billing, and led end-to-end delivery of Cloud Portal 2.0.
The scope: a shared developer platform bringing provisioning, access, and cost management into the engineering workflow.
Shared architecture, reusable platform work, and the people who operate it.
Architecture alignment
Make decisions reusable.
At Walmart, I established architecture governance and distributed systems training, alongside query normalization and tokenization adopted across Search teams.
At IBM, I led a cross-functional team of 10 engineers responsible for monitoring and alerting, and served as the subject-matter expert for L1 escalations.
From secure applications to global infrastructure, developer platforms, and ML serving. Technical leadership grounded in operating the systems I build.
2024 to presentWalmartStaff Software Engineer
2023 to 2024PayPalStaff Software Engineer
2019 to 2023FordSenior Software Engineer
2016 to 2019IBMSoftware Engineer / Team Lead
2015MicrosoftService Engineer Intern
2012 to 2014InfosysSenior Systems Engineer
MS in Computer Science, UT Dallas. MBA, University of Illinois Urbana-Champaign.
A local control plane for running and supervising coding-agent sessions across Git worktrees.
Engineering focus: separate workspaces and honest operational visibility. Worktrees separate file changes, the local dashboard requires a capability token, and unavailable usage measurements remain unknown.