Eric Smith

- Architected 16-stage multi-agent orchestration in TypeScript on GitHub Actions where 26 tool-augmented AI agents autonomously produce security-scanned, schema-validated pull requests, enforcing policy-as-code risk classification, 7 verification gates, prompt injection detection, and ISO-8601 compliance audit trails. Automated per-issue ROI calculation tracking time savings, risk reduction value, per-agent cost breakdown, and token usage with research-backed citations, eliminating manual survey-based estimation. - Engineered an intelligent model router with 4-level reasoning classification and a 3-tier provider cascade (GitHub Copilot SDK, GitHub Models REST API) selecting from 50+ LLMs across 6 providers with stage-criticality multipliers. Designed a Best-of-N competition framework where 3 LLMs generate competing plans scored by a deterministic evidence-weighted ranker. - Designed a bidirectional Azure DevOps-to-GitHub orchestration bridge using a "Mirror Issue" pattern where WIQL-based polling creates mirror GitHub Issues from ADO work items via repository_dispatch, triggering full AI orchestration without external middleware. Implemented idempotent deduplication, automated state transitions across the work item lifecycle, and structured summary postback to ADO Boards. - Built a composable agent platform with a YAML-frontmatter workflow system compiling natural language agent definitions into governed workflows, a Handlebars prompt builder allocating per-agent token budgets across 12 auto-discovered skills, and 24 JSON Schema contracts validated by AJV with automated repair loops, enabling non-engineers to author production-grade AI workflows. - Instrumented self-improving observability tracking per-agent cost telemetry, evidence quality, and human edit rates, generating refinement backlog items automatically. Enforced a 3-level emergency brake system and agent separation-of-concerns where no agent can bypass another's security checks.

Engineered AI-powered duplicate detection system using Claude AI (AWS Bedrock) to automatically identify and cancel duplicate triage tickets with ≥90% confidence thresholds. Implemented smart categorization across seven ticket types with safety protections, automated GitHub Actions workflows, and Slack notifications

- Designed and implemented Node.js function-as-a-service microservice using TypeScript / AWS SDK for JavaScript by migrating three monolithic Python codebases. Leveraged hexagonal architecture, domain-driven design, and facade APIs to increase revenue reporting accuracy from 60 to 99.9% for 100 B2B customers and reduce team on-call work by 50%, while migrating features to the new microservice and improving existing monolithic applications - Directed cross-functional collaboration across ten engineering squads to build two RESTful and GraphQL APIs using AWS AppSync, API Gateway, Lambda, Kinesis, SQS, and S3, driving the launch of new product features for 100 B2B customers. Engineered bulk / paginated API endpoints with caching and event-driven architecture, optimizing for scalable and highly accurate communication between multiple backend and micro-frontend services - Architected two Typescript AWS Batch computing services using Node.js, Docker, and EC2 to execute 200 scheduled and custom jobs, replacing failing Python Lambda services and enhancing scalability of critical product functionality for 90 B2B customers. - Increased feature development speed by 20% through strategic use of GitHub Copilot and ChatGPT Enterprise, improving code quality and enabling rapid, high-impact delivery of software features - Implemented Datadog tracing, monitoring, and alerting across three services to achieve high visibility into performance, enabling proactive issue detection and enhancing overall system reliability - Created 4 AWS Glue ETL data pipelines using Python / PySpark and a bitemporal PostgreSQL database supported by MirkoORM data mapping to achieve company-wide data product reporting to improve revenue forecasting

- Developed Node.js function-as-a-service microservice in TypeScript with AWS SDK for JavaScript, transforming three Python monoliths into a domain-driven, loosely coupled service to unify financial data ingestion products. Led technical design initiatives to standardize financial data quality for 75 B2B customers, reducing engineering on-call work by 15% - Deployed two CI/CD pipelines with Github Actions / Shell scripts to identify changed resources and run all unit, integration, and service contract tests, saving AWS resource costs and reducing deploy times by 10-50% per pull request

- Created client data validation system with Python, AWS Cloud Services, and Google / Snowflake / Slack APIs, identifying $1 million in unreported revenue annually and saving 75 hours of manual work per month for 3 teams - Designed and built KPI product and data validation dashboards using Python, Google API / Data Studio, and Jira data, enabling engineering and product leadership to identify and reduce on-call costs by 10% annually - Conducted technical product launch and training of 3 data pipeline suites for 10 B2B customers, activating the company’s marketplace payment and data architecture objectives for these customers and increasing their revenue by 15% annually

• Collaborated with 7 science teachers to design and facilitate objective-driven lessons for over 195 students, resulting in 75% mastering curriculum standards, exceeding department goal by 5% • Designed differentiated lessons for 55 English Language Learners and students with academic modifications, culminating in 91% completing year on grade level • Led 10-student FIRST Robotics team in learning and implementing systems engineering design and Java programming concepts to build robot for two regional competitions

• Supported 41 Kode With Klossy students build 4 iOS applications by leading 2-week Swift and GitHub programming courses

• Selected amongst 48,000 applicants nationwide with an 11% acceptance rate to join a diverse network of leaders who commit two years to teach in a low-income community public school • 2018 TFA-DFW corps member leading physics and computer science curriculum design and instruction at L.G. Pinkston High School.

• Managed deliverable development process for 10 federal business proposals with 100% compliance using Trello • Forecasted work timelines for team of 6, designed proposal documents, managed quality assurance, and delivered proposals to clients • Performed 5 subject-expert interviews and customer research to support completion of venture capital study helping U.S. Navy reduce ship repair times by 10% through digital logistics solutions

• Designed, programmed, and implemented 4 capacity analysis interactive dashboards in Excel, reducing production employee overtime hours by 15% • Created data visualizations using pivot tables, macros, and VBA to support key business decisions • Collaborated with 15 engineers to reduce engine build times by 10% through manufacturing process redesign

• Developed web application in R to enable Operations project managers to purchase and use fewer energy-intensive products • Designed interactive web dashboard to host product life-cycle analysis program using R Shiny, HTML, and CSS • Wrote 6 NASA Feature stories and 4 newsletters read by 800+ employees via creating WordPress website content and conducting 6 employee interviews • Partnered with Education Office to design sustainability curriculum for NASA high school summer programming

• Increased school clientele by 150% over 2 years by designing and implementing new music education curriculum • Led 10 professional development clinics for over 500 attendees, resulting in hiring of 10 new employees • Spearheaded bi-weekly professional development sessions and daily collaboration with 15 teachers

Online Computer Science Post-Baccalaureate Degree