OTCnet — AI Implementation & Governance

• —Lead engineer for the AI-in-SDLC track within OTCnet's five-track AI governance program (AI-in-SDLC, reports/dashboards, insights/trends, chatbot, adaptive training); own engineering breakdown, task assignment, and stakeholder reporting in recurring review cycles.
• —Authored the program's AI risk-assessment and trade-off scoring framework — scoring use cases on data confidence and relevancy, model and algorithm suitability, algorithm attributes (e.g., decision-tree transparency), lifecycle, equity and inclusion, and privacy. Delivered scored technical assessments and ran proof-of-concept evaluation in the governance sub-council.
• —Built an AI-assisted Section 508 accessibility triage layer: Azure DevOps pipeline runs Playwright + axe-core against OTCnet, findings and screenshots land in Azure Blob, Azure Functions orchestrate Azure OpenAI calls (GPT-4o for multimodal triage, GPT-4.1 for long-context batch, o3-mini for escalated reasoning) under RAG against an Azure AI Search index of OTCnet component patterns, known false positives, and 508 techniques.
• —Triage output — severity and false-positive classification, OTCnet-component mapping, and VPAT-ready remediation language — routed as Azure DevOps work items to the accessibility queue; human testers review and dispose, dispositions roll to Power BI. AI never makes the final compliance call.
• —Deployed in Azure Government (FedRAMP High) inside a private endpoint within the OTCnet VNet; designed to reduce false-positive and duplicate burden on a small accessibility team by ~40%.
• —Prototyped an AI drafting accelerator for SSP control narratives and assessor-response support — Azure OpenAI GPT-4.1 with hybrid keyword + vector retrieval and semantic ranking over fiscal-service control mappings, prior approved narratives, and OTCnet system docs. Compresses first-draft time on ATO-adjacent artifacts so ISSOs and security SMEs spend time on judgment review; all narratives stay SME-reviewed.
• —Challenged leadership's 40–60% time-savings projection with an independent cost-benefit analysis grounded in realistic adoption curves and review overhead; reframed the initiative around sustainable adoption over inflated efficiency claims.

OTCnet — Test Automation & Performance

• —Expanded test ownership from manual coverage into full automation and performance testing across OTCnet's federal payment workflows.
• —Building an automated suite with Python, JavaScript, Selenium, and Playwright — semantic locators and self-healing scripts cut maintenance overhead from UI churn.
• —Performance testing via Azure Load Testing with AI-assisted JMeter script generation for realistic load profiles.

OTCnet — Manual Test Engineering

• —Owned manual test coverage for OTCnet's federal payment workflows under CitiBank-managed delivery.
• —Executed functional and regression testing to ensure production stability; identified edge cases and managed the defect lifecycle from discovery to verification.

Jira → Azure DevOps Migration

• —Led the technical migration of project management infrastructure from Jira to Azure DevOps for a high-visibility federal program.
• —Architected the mapping of complex work item hierarchies and automated workflows to preserve data integrity across the transition.

DoE — Cloud Infrastructure (Kubernetes)

• —Contributed to Kubernetes-based cloud migration for Department of Education workloads across containerized microservices.
• —Built declarative infrastructure-as-code with Terraform and AWS CDK; implemented idempotent deployment patterns for consistent system state.
• —Helped deliver the Kubernetes-based MVP enabling scalable, compliant, cloud-native deployment inside Booz Allen's Trusted Environment; contributed to VPC setup and CI/CD pipelines.

DoE — FAFSA AI Assistant

• —Designed and built an AI-powered navigation agent to help students complete FAFSA applications — targeting high support volume and low-income aid-access bottlenecks.
• —React frontend, Flask backend, GPT-3.5 via OpenAI API, and a LangChain RAG architecture over federal financial aid documentation.
• —Engineered custom state management and prompt templates to stabilize non-deterministic LLM workflows and improve response reliability for high-volume inquiries.
• —Won the firm-wide Booz Allen AI innovation competition (200+ attendees including senior leadership); awarded dedicated R&D engineering hours. Development paused when the DoE contract restructured.