Resume

• Architected, built, and deployed a production ITSM SaaS platform — 7 NestJS microservices with SAML SSO, async job queues, and RAG-based ticket routing — replacing a licensed commercial product at The College of Wooster. Serves 4,000+ active users in a community of 35,000.
• Identified the market gap by analyzing 500+ historical tickets and interviewing 16 IT staff; secured executive buy-in and unanimous approval from 4 divisional directors to replace the legacy TopDesk platform.
• Integrated SAML 2.0 SSO with Microsoft Entra ID for institutional compliance. Built a local RAG pipeline (pgvector + BGE embeddings + Gemini 2.5 Flash) that reduced ticket creation time by ~87.5%.
• Load-tested with k6 across smoke, spike, stress, and soak profiles: sustained P95 latency under 20ms at 50 concurrent users over a 30-minute soak; identified capacity ceiling at 500 VU for next-iteration hardening.
• Shipped CI/CD via GitHub Actions (smoke, unit tests, container build, registry push on every commit); coordinated on-prem deployment on SUSE Linux with campus system administrators.
• Achieved 40% ticket, 70% reservation, and 100% asset adoption in two weeks with no formal training sessions. Estimated $33.5K annual savings ($20K license retirement + $13.5K operational efficiency).

• Resolved 500+ support tickets with 90% user satisfaction — the operational insight that led to designing and building AtlasDesk.
• Built a ticket-classification system using MiniLM embeddings for automatic routing across Wooster-specific categories.
• Automated OS installation and AD domain binding for 200+ lab computers via PowerShell, cutting a 4–5 hour weekly task to 10 minutes (96% reduction).
• Leading the AtlasDesk pilot integration into production IT infrastructure, coordinating with 12 IT staff and 14 student technicians on migration from legacy TopDesk.

• Prototyped an LLM-driven co-pilot (LlamaIndex, FastAPI) over 250,000+ internal workflows, enabling natural-language queries on complex energy data.
• Designed a Model Context Protocol client–server architecture integrating 15+ external tools via automatic schema discovery.
• Contributed to a team benchmarking 13 LLMs (GPT-4, Claude, Llama) across single- and multi-agent architectures using GPT-4-as-judge evaluation.

• Optimized 30+ Django REST APIs and PostgreSQL queries, lifting Lighthouse performance from 72% to 96%.
• Contributed to Stripe payment integration and implemented JWT with rotating refresh tokens across 10,000+ user accounts.

• AI GitHub App automating code review on a 30,000+ line codebase via live webhooks and line-specific PR comments.
• PostgreSQL cache keyed on file-content hashes cut review latency 88% (10.1s → 1.2s) and per-review cost from $0.12 to $0.014.

• Identified a V2X "routing-loop" failure mode in autonomous-vehicle coordination where inter-vehicle communication degraded performance.
• Designed an Object Memory Management framework that reduced coordination delay by 75.7% and wait time by 88%. Presented at NCUR 2025.

Departmental Honors in Computer Science. Co-Lead, Google Developer Student Club. Stanford TreeHacks 2024 Participant.