• Led the network repair operating model, shifting network operations from reactive ticket triage to structured diagnosis workflows, increasing actionable repairs from 10% to 50% and establishing the standard across all datacenters

• Designed and deployed an AI-native ticket routing system, replacing 1,200 lines of code and improving accuracy from 67% to 95%, including a self-improving pipeline that proposes new rules, and opens its own code reviews

• Delivered repair automation across 68K+ devices, 40K+ optics, and 130K memory replacements, reducing maintenance cycles and avoiding $4.5M in operational labor costs

• Built Model Context Protocol (MCP) tools and a shared diagnostics library powering AI diagnostic agents across platforms

• Established the repair platform health scorecard with concrete SLOs adopted across adjacent teams

• Mentored engineers and expanded team ownership into AI, datacenter, and backbone network domains

• Program and technical lead for a 20 person engineering team across 8 workstreams focusing on automation, achieving 21% reduction in repair time, 39% improvement in human efficiency, and supporting 169% network growth

• Led cross-team initiatives aligning systems and workflows, establishing repeatable automation across infrastructure teams

• Defined architecture and interfaces for next-generation repair platform integrating diagnosis, orchestration, and execution services

• Mentored engineers and expanded team capability in automation and reliability engineering

• Consistently received top performance ratings (“Greatly Exceeds” / “Redefines Expectations”)

• Improved event routing accuracy and reduced misclassified repair events by resolving systemic gaps across the repair workflow

• Cut alert noise by 50–55% and reduced 48-hr drain issues by 95% through automated suppression and detection logic

• Authored oncall guidelines, creating standardized playbooks for rapid troubleshooting and operational readiness

• Helped scale operations to support 17k+/yr external repair inquires without increasing headcount by improving workflows and reducing mean-time-to-repair by ~35%

• Reduced outbound ISP ticket noise by 88% by redesigning the automation with data-driven delay logic

• Built and consolidated multiple dashboards, used for ticket auditing, operations, and identifying automation opportunities

• Optimized global CDN performance by tuning routing across POPs, reducing latency and improving service quality for end users

• Automated configuration deployment (Python, GitHub, Ansible), increasing repeatability and reducing operational overhead

• Evaluated and expanded peering and transit capacity, supporting Fastly’s global traffic growth