cover_letter / art_5ukL4VyGE_k

Dear Mesh Hiring Team, Mesh is tackling one of the most consequential infrastructure problems of this decade: making tokenized assets actually spendable in the real world. The gap between trillions of dollars in on-chain value and the ability to use that value for everyday commerce is not a UX problem — it is an orchestration and trust problem, and it is exactly the kind of problem I have spent the last several years building toward. When I read that Mesh needs someone to figure out how agents will buy, sell, and move money before a product exists to ship, I recognized the work immediately — because I have been doing a version of it. **Technical and AI Foundation** My most directly relevant build is OpenClaw, a multi-agent orchestration framework I designed and shipped inside StreamIO AI. OpenClaw implements a gateway protocol with subagent delegation, profile management, and session switching — coordinating AI agent workflows across real estate, insurance, health/dental, and financial-markets verticals. The core design problem OpenClaw solves is the same one agentic commerce surfaces: how does an agent take an action with real-world consequences (submitting a CMA report, triggering a market data pull, initiating a transaction) without a human in the confirmation loop, while still maintaining auditability and control? I did not read about this problem; I wired it up, watched it break, and rebuilt it. On the financial side, I built Fintellect — a mobile-first AI investing platform — with a RAG retrieval pipeline (ChromaDB), multi-provider LLM orchestration across Claude, GPT-4, and Gemini with fallback routing, and a paper trading engine on live Alpaca market data supporting market, limit, stop, and stop-limit orders with P&L tracking and AI portfolio analysis. I also hold a Topstep Funded Trader certification, which means I have operated under real risk constraints, not just simulated ones. Genuine curiosity about how money moves is a requirement in your JD — I would describe it as a preoccupation. My AI/ML foundation goes deeper than product. I published at NeurIPS 2014 on neural networks for protein structure prediction, hand-coded BPTT in C++ at Berkeley in 2004, and in 2026 rebuilt that system in PyTorch spanning 413 parameters to 8B — a 19-million-fold scale increase. I built a post-training RL workbench benchmarking GRPO, DPO, PPO, DAPO, and eight other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL with live SSE metric streaming. I understand the model layer, not just the API surface. **Why This Role** My arc — from platform infrastructure at Intuit (675M+ engagements, 50K TPS) to founding AI companies where I write the code, ship the product, and do the customer discovery — maps directly onto what Mesh is asking for: someone who can think in systems, build working prototypes on live rails, and operate without a product brief. The Founding PM for Agentic Commerce role is not a management job; it is an exploration and proof-of-concept job with a PM title. That is the job I have been doing. **What Excites Me About This Specific Role** The framing in your JD — "thinking without doing is just speculation; doing without thinking is giving in to FOMO" — is the clearest articulation I have seen of what separates useful AI product work from noise. I am drawn to the explicit mandate to embed in protocol working groups and standards bodies, because the agent payment space will be shaped by whoever shows up early and earns credibility before the standards calcify. The instruction to recruit experiment customers from Mesh's existing base rather than waiting for a formal GTM motion is also exactly how I have operated — at Fintellect and Vantage, I ran customer discovery interviews and iterated directly from trader and job-seeker feedback before any formal launch. **Selected Prior Experience** - Designed and shipped OpenClaw multi-agent orchestration framework: gateway protocol, subagent delegation, profile management, and session switching — coordinating AI agent workflows across real estate, insurance, health/dental, and financial-markets industries. - Built Fintellect's AI investing platform with live Alpaca market data, paper trading ($100K virtual portfolio with market/limit/stop/stop-limit orders, P&L, and AI portfolio analysis), and 13 specialized domain AI advisors with RAG retrieval and multi-provider LLM fallback routing. - Shipped Vantage with 40+ agent tools over a FastAPI backend, enforced mutation approvals, streaming agent chat (SSE), and async job polling with idempotent endpoints — production agentic architecture, not a demo. - Scaled Intuit's 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. - Built aeval, a local-first AI model evaluation platform with adversarial safety testing, bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, and CI/CD regression detection — because I evaluate what I ship. - 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. - NeurIPS 2014 published researcher; hand-coded BPTT in C++ (2004); 2026 PyTorch rewrite spanning 413 to 8B parameters — technical credibility that runs through the full stack. **Closing** Mesh's mission — enabling consumers to pay and be paid with any asset — only becomes real when the infrastructure can handle principals that are not human. Agents that can transact autonomously, with auditability and without friction, are not a feature on top of Mesh's core; they are the next version of what Mesh is. I want to be the person who figures out what that looks like, builds the first working version of it on live rails, and makes the case for where Mesh bets. I would welcome the chance to walk through something I have built and the decisions inside it. Sincerely, **O. Felix Amoruwa** famoruwa@berkeley.edu · 909-731-9011 · felixamoruwa.info