cover_letter / art_Tl5t4Tnvhww

Dear OpenRouter Hiring Team, OpenRouter sits at a genuinely important inflection point: as enterprises race to operationalize LLMs, the infrastructure layer that routes, manages, and optimizes across providers without lock-in is becoming load-bearing for the entire industry. That problem — building the plumbing that lets the best AI teams move fast and stay future-proof — is exactly the kind of platform challenge I have spent 12 years preparing for. My path from hand-coding BPTT in C++ at UC Berkeley to scaling Intuit's developer platform to 675M+ engagements and building production AI infrastructure from scratch as a founder makes this role feel like a natural next chapter, not a lateral move. ## Technical Foundation My platform engineering credibility is grounded in systems work at scale. At Intuit, I owned the ICE platform — the infrastructure layer powering QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma. I drove a rSocket migration that scaled throughput from 6K to 50K TPS while supporting approximately 1.5M concurrent connections at sub-25ms TP99 latency. That is the kind of distributed systems trade-off — protocol selection, connection management, tail latency — that maps directly to operating a high-throughput LLM routing layer. I also led the Mailchimp GCP-to-AWS migration for MSaaS, delivered a Java-based drift detection library that scanned Git repos for configuration drift across a large microservice estate, and built Asterias, a declarative asset lifecycle management platform with a GraphQL API — all while managing a multi-team roadmap and reporting to senior engineering leadership. At Splunk, I owned Search Service (Go microservices), Search Catalog (PostgreSQL metadata), and SPL/SPL2 — delivering the Scheduler Service end-to-end in roughly four months and achieving up to 10x query performance improvements for a beta Fortune 500 customer through benchmark-driven optimization. At Kaiser Permanente, I built and scaled Splunk Logging-as-a-Service to 1.7 TB daily volume across 200+ internal enterprise customers, and introduced Redis-based caching across the enterprise to address fault tolerance and data redundancy at scale. On the AI infrastructure side, I built a production RL post-training workbench that benchmarks GRPO, DPO, PPO, and nine other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL — with GPU Docker passthrough, live SSE metric streaming, and standardized throughput/memory/convergence benchmarking. I also built aeval, a local-first model evaluation platform with a FastAPI orchestrator, TimescaleDB, Redis job queue, and Next.js dashboard. These are not toy projects; they are systems I built to solve real evaluation and training infrastructure gaps, and they reflect how I think about platform problems: composable, observable, and built for iteration. ## Why This Role OpenRouter's platform team owns the infrastructure that routes millions of LLM requests daily across a multi-provider, multi-model surface area — with edge compute, cloud infrastructure, and data systems all in scope. That combination of distributed systems depth, developer-facing platform ownership, and the need to make fast architectural calls in a lean org is where I have consistently done my best work. The fact that this role reports directly to the CEO and requires both technical hands-on contribution and team leadership matches how I have operated: at Intuit I was a Staff PM who wrote Java libraries, ran SQL against BigQuery, and presented language investment strategy to the CTO in the same week. I am particularly drawn to the edge and routing infrastructure challenges OpenRouter faces. Optimizing request routing across providers — balancing latency, cost, capacity, and model quality — is a systems problem with real product consequences, and it is one where architectural decisions compound quickly. I want to be in the room where those calls are made, and I want to be writing the code that validates them. ## Selected Relevant Experience - **Scaled ICE platform to 675M+ engagements in FY23** across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma; drove rSocket migration from 6K to 50K TPS supporting ~1.5M concurrent connections at sub-25ms TP99. - **Delivered ICE Self-Service platform** (DevPortal, GitOps config, ICE Playground), reducing developer onboarding from 2–3 weeks to minutes in pre-prod and under 24 hours for production, while mitigating $1M+ in projected opex growth. - **Led Mailchimp GCP-to-AWS migration** for MSaaS, delivering Golang service template, MySQL persistence integration, and updated DevPortal documentation to meet production deadline. - **Built and shipped Splunk Scheduler Service end-to-end in ~4 months**, enabling scheduled search for first-party applications; led query optimization achieving up to 10x performance improvements for a Fortune 500 beta customer. - **Built RL post-training workbench** benchmarking 12 algorithms (PPO, GRPO, DAPO, DPO, and others) across TRL, VeRL, OpenRLHF, and NeMo RL — with GPU Docker passthrough, live SSE metric streaming, and cross-framework throughput/memory/convergence benchmarking. - **Built aeval evaluation platform** with FastAPI orchestrator, TimescaleDB, Redis job queue, bootstrap confidence intervals, Welch's t-test, and automated CI/CD safety gates — a production-grade observability and regression detection system for model quality. - **Extended Java and Python SDK Starter Kits** with scaffolding templates, Gradle/Maven build configurations, testing frameworks, and CI/CD integration — enabling developers to go from zero to production-ready microservice in minutes. - **Conducted enterprise-wide Service Language Assessment** across 9 languages, analyzing usage data and developer feedback to inform strategic investment decisions presented to the CTO — the kind of cross-cutting platform analysis that informs where to invest and where to cut. ## Closing OpenRouter's mission — giving the world's most advanced AI teams the flexibility to move fast, scale confidently, and stay future-proof — is one I want to contribute to directly. The infrastructure layer that makes that possible is not a commodity; it requires engineering leadership that can stay hands-on while making the architectural calls that keep a lean team moving at the pace the business demands. That is the kind of leader I am, and I would welcome the opportunity to bring that to OpenRouter. Thank you for your consideration. --- **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info