cover_letter / art_Rq6VeYKVmF4

Dear Exa Hiring Team, Exa is doing something genuinely hard: building a search engine that meets the demands of the AI era — one where agents, not humans, are often the primary consumers of retrieved information. That distinction matters enormously, and it shapes every product decision from latency to relevance to API ergonomics. I came to appreciate this acutely while building the OpenClaw multi-agent orchestration framework at StreamIO, where the quality of information retrieval was the single largest determinant of whether an agent workflow produced a useful output or a hallucinated one. That experience made me a close follower of Exa's approach, and it's what draws me to this role. **Technical and AI Foundation** My technical grounding runs from the foundational to the current. In 2004 I hand-coded a backpropagation-through-time neural network in C++ for protein structure prediction at UC Berkeley — work that was accepted at NeurIPS 2014. In 2025–2026 I rewrote that system in PyTorch, scaling from 413 parameters to 8B (a 19-million-fold increase), adding five neural architectures, MLflow experiment tracking, Optuna hyperparameter optimization, and FastAPI serving across six Docker containers. That arc — from first principles to production-scale ML — gives me the ability to read model specifications, evaluate framework tradeoffs, and have credible technical conversations with engineering teams rather than around them. More recently, I built a 3-phase RL post-training workbench benchmarking GRPO, DPO, PPO, DAPO, and eight other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL, with live SSE metric streaming and GPU Docker passthrough. I also built aeval, a local-first model evaluation platform with bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, and adversarial safety testing — a FastAPI orchestrator backed by TimescaleDB, Redis, and Ollama. These projects were not research exercises; they were built to inform real product decisions about which training and evaluation approaches are worth productizing. On the developer platform side, at Intuit I owned the ICE platform from roadmap through execution — extending Java and Python SDK Starter Kits, delivering a self-service DevPortal with GitOps configuration, and reducing developer onboarding from two to three weeks down to minutes in pre-production. That platform scaled to 675M+ engagements in FY23 across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma, with throughput growing from 6K to 50K TPS via an rSocket migration supporting ~1.5M concurrent connections at sub-25ms TP99. **Why This Role** My career has moved steadily toward the intersection of developer-facing platforms and AI-native products — from enterprise SDK infrastructure at Intuit to founding two AI companies where I was simultaneously the PM, the engineer, and the first customer. Exa sits exactly at that intersection, and the specific challenge of extending a core search API to solve new use cases — with 50% of the work happening directly alongside the customers integrating it — is the kind of product work I find most generative. What excites me most about this role is the customer-proximity model. Watching AI-native startups and enterprises actually use the Exa tool, identifying where they get stuck, and converting that friction into productized capability is precisely how I built the Vantage job-search platform and the Fintellect financial-education app — both shipped from zero to App Store with iterative customer discovery driving every prioritization decision. At Exa's scale and with its customer base, that same loop produces compounding returns across the entire API platform. **Selected Relevant Experience** - **Developer platform at scale:** Delivered ICE Self-Service platform (DevPortal, GitOps config, ICE Playground) at Intuit, reducing developer onboarding from 2–3 weeks to minutes; scaled to 675M+ engagements in FY23 and mitigated $1M+ in projected opex growth. - **SDK and tooling ownership:** Extended Java and Python SDK Starter Kits with scaffolding templates, build configurations (Gradle/Maven), testing frameworks, and CI/CD integration — enabling developers to reach production-ready microservices in minutes. - **Multi-agent API architecture:** Implemented OpenClaw multi-agent orchestration framework (gateway protocol, subagent delegation, profile management, session switching) coordinating AI agent workflows across real estate, insurance, health/dental, and financial-markets verticals. - **RAG and LLM orchestration:** Architected a RAG retrieval pipeline (ChromaDB vector store) with multi-provider LLM orchestration (Claude, GPT-4, Gemini), fallback routing, structured-output validation, and token-budget optimization for Fintellect's 13 specialized AI advisors. - **0-to-1 developer product with 40+ agent tools:** Shipped Vantage on iOS and macOS with a shared FastAPI backend exposing 40+ agent tools, enforced mutation approvals, streaming agent chat via SSE, and MCP servers for screen capture and job search. - **Customer discovery driving prioritization:** Led customer discovery interviews and iterative product refinement across both StreamIO and Fintellect, using trader and job-seeker feedback to sequence feature development and go-to-market execution. - **Search and query performance:** At Splunk, owned Search Service (Go microservices), Search Catalog (PostgreSQL metadata service), and SPL/SPL2; led a query performance optimization initiative achieving up to 10x improvements in Splunk Cloud Services search for a beta Fortune 500 customer. **Closing** Exa's mission — making the web's knowledge reliably accessible to the agents and applications that need it — is one of the more consequential infrastructure bets in AI right now. The quality of what agents can do is bounded by the quality of what they can retrieve, and getting that right at API scale requires product thinking that is simultaneously technical, customer-proximate, and strategically disciplined. That is the work I have been building toward, and I would welcome the opportunity to bring it to Exa. Thank you for your consideration. --- **O. Felix Amoruwa** famoruwa@berkeley.edu · 909-731-9011 · felixamoruwa.info