cover_letter / art_GvU7d9vlPtQ

Dear Lyft AI Platform & Marketplace Applications Hiring Team, Lyft's purpose — to serve and connect — is realized through the infrastructure that no rider ever sees: the marketplace algorithms, AI platforms, and decision systems that match supply to demand in real time across millions of trips. That infrastructure is exactly the kind of work I have spent the last several years building, from the ground up. When I read the description of the Central Market Management and Applied Intelligence org, I recognized the shape of the problem immediately: a platform team whose value is only realized when the rest of the company trusts it enough to build on it. ## AI/ML and Technical Foundation My technical foundation in AI is not recent. In 2004 I hand-coded a neural network in C++ with custom backpropagation through time (BPTT) for protein secondary structure prediction — work that was accepted at NeurIPS 2014. In 2026 I rewrote that same system in PyTorch, scaling from 413 parameters to 8 billion across five architectures (feedforward, GRU, Transformer, ESM-2, multi-task), with MLflow experiment tracking, Optuna hyperparameter optimization, and FastAPI serving — a 19-million-fold increase in parameter scale across two decades of personal research. More directly relevant to this role: I built a production RL post-training workbench that benchmarks GRPO, DPO, PPO, DAPO, REINFORCE, REINFORCE++, RLOO, SimPO, IPO, KTO, ORPO, and SPPO — twelve algorithms in total — across four training frameworks (TRL, VeRL, OpenRLHF, NeMo RL) with GPU Docker passthrough, live SSE metric streaming, and standardized throughput/memory/convergence benchmarking. I built aeval, a local-first model evaluation platform with bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, adversarial safety testing with refusal detection, and CI/CD regression gates — running on FastAPI, TimescaleDB, Redis, and Ollama. And I designed and shipped OpenClaw, a multi-agent orchestration framework with a gateway protocol, subagent delegation, profile management, and session switching — the same architectural pattern that underpins modern agentic systems at scale. This is not familiarity with AI tooling. It is hands-on, production-grade implementation across the full stack: reward modeling, post-training RL, multi-agent orchestration, and evaluation infrastructure. ## Why This Role The through-line of my career has been building platforms that other builders depend on — developer frameworks at Intuit, search infrastructure at Splunk, enterprise logging services at Kaiser — and then driving the adoption that makes those platforms real. The Lyft AI Platform role sits at exactly that intersection: technical depth to earn credibility with engineers and scientists, and the product and communication skills to translate platform capabilities into business outcomes for non-technical leadership. What specifically draws me to this role is the scope of the agentic AI mandate. Lyft's marketplace is a real-time, multi-sided system where decision latency and decision quality are both first-order concerns. Building an agentic AI platform that accelerates decision-making across hundreds of internal applications — and then evangelizing adoption of that platform across the company — is a product challenge I am well-positioned to take on. The JD's emphasis on "context engineering" and "multi-agent orchestration" maps directly to the systems I have designed and shipped. ## Selected Prior Experience - **RL Workbench (2026):** Built a 3-phase post-training RL platform covering Reward Lab (A/B testing reward functions across GSM8K, MATH, HumanEval, UltraFeedback), a TRL-powered training Playground with live metric streaming, and an Arena for head-to-head framework benchmarking (TRL, VeRL, OpenRLHF, NeMo RL) — directly applicable to evaluating AI technology bets and guiding build-vs-buy decisions. - **OpenClaw Multi-Agent Orchestration (Streamio AI, 2024–Present):** Designed and implemented a production multi-agent orchestration framework with gateway protocol, subagent delegation, profile management, and session switching — enabling coordinated AI agent workflows across multiple industry verticals. - **aeval Evaluation Platform (2025–2026):** Built a rigorous model evaluation platform with statistical testing (bootstrap CI, Welch's t-test, Cohen's d), adversarial safety gates, and CI/CD integration — the kind of measurement infrastructure needed to track KPIs and report platform progress to leadership with credibility. - **ICE Platform at Intuit (2021–2024):** Delivered the 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. Achieved 275% YoY growth in ICE engagements, scaling to 675M+ in FY23; scaled throughput from 6K to 50K TPS via rSocket migration supporting ~1.5M concurrent connections at sub-25ms TP99. - **Developer SDK and Platform Infrastructure (Intuit, 2021–2024):** Extended Java and Python SDK Starter Kits with scaffolding templates, build configurations, testing frameworks, and CI/CD integration; conducted enterprise-wide Service Language Assessment across 9 languages, presenting findings to the CTO — demonstrating the ability to synthesize technical data into executive-level strategy. - **Splunk Search Orchestration (2019–2021):** Owned Search Service (Go microservices), Search Catalog (PostgreSQL metadata service), and SPL/SPL2 — delivered Scheduler Service end-to-end in ~4 months, demoed at .conf19, and achieved up to 10x query performance improvements for a Fortune 500 beta customer. Designed RICE-based prioritization framework across 3 microservice backlogs. - **RAG and LLM Orchestration (Fintellect AI, 2024–Present):** Architected a RAG retrieval pipeline with ChromaDB vector store, multi-provider LLM orchestration (Claude, GPT-4, Gemini) with fallback routing, structured output validation, and token budget optimization — directly relevant to the AI platform architecture Lyft is building. ## Closing Lyft's mission to serve and connect is only as good as the systems that make each connection reliable, efficient, and fair. The AI platform org is the team that determines whether Lyft's most critical decisions are made well — and whether the rest of the company has the tools to make them faster. I have spent my career building exactly these kinds of foundational platforms, earning the technical credibility to work alongside engineers and scientists, and developing the communication and strategy skills to bring non-technical leadership along. I would welcome the opportunity to bring that combination to Lyft. Thank you for your consideration. **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info