cover_letter / art_Iv8tHRmAWxo

Dear Cresta Hiring Team, Cresta is doing something genuinely important: turning every customer conversation into a measurable business outcome by combining conversational AI agents, real-time augmentation, and conversation intelligence into a single platform. The companies deploying Cresta — United Airlines, Cox Communications, Marriott — are exactly the kind of enterprises where AI agent quality and deployment rigor translate directly into millions of dollars of value or loss. My interest in this role is grounded in direct experience: over the past year I have designed, built, and deployed multi-agent AI systems from scratch — not as a PM watching engineers build, but as the person writing the orchestration logic, the agent delegation protocols, and the evaluation harnesses myself. **Technical and AI Foundation** My AI work spans two decades and is hands-on throughout. In 2004 I hand-coded a neural network in C++ with custom backpropagation through time for protein structure prediction — work that was accepted at NeurIPS 2014. In 2025–2026 I rebuilt that system in PyTorch across five architectures (feedforward, GRU, Transformer, ESM-2, multi-task), scaling from 413 to 8 billion parameters. More directly relevant to Cresta: I built OpenClaw, a multi-agent orchestration framework with a gateway protocol, subagent delegation, profile management, and session switching — enabling coordinated AI agent workflows across real estate, insurance, health/dental, and financial markets verticals. This is the same class of architecture Cresta deploys in contact center environments, and I built it end-to-end. I also built aeval, a local-first model evaluation platform covering factuality, reasoning, instruction-following, safety, and code generation — with bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, and automated safety gates integrated into CI/CD. Rigorous evaluation of agent behavior before and after deployment is exactly the discipline a Forward Deployed PM at Cresta needs to own, and I have built the tooling to do it. My RL Workbench benchmarks GRPO, DPO, PPO, DAPO, and eight other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL — giving me a working understanding of how post-training decisions shape model behavior in production, which matters when optimizing AI agents against real customer outcome metrics. **Why This Role** My arc — from NeurIPS researcher to Staff PM scaling platform infrastructure to 675M+ engagements at Intuit, to founding and building AI agent products — points directly at what Cresta's Forward Deployed PM role requires: the ability to sit with a senior customer executive, scope an AI agent deployment, design the workflows, and then roll up my sleeves and build alongside the engineering team. I have done each of those things, and I want to do them together in a high-stakes, customer-facing context where the outcomes are visible and the feedback loop is tight. **Role-Specific Connection** What excites me most about this role is the full deployment lifecycle ownership — from pre-sale scoping through post-go-live optimization — combined with the expectation that the PM is a builder, not just a coordinator. Cresta's platform sits at the intersection of conversational AI agents and real-time human augmentation, which means the agent design decisions I make will affect both automation rates and the live experience of human agents working alongside AI. That dual accountability — to automation metrics and to human experience quality — is a more interesting design problem than pure automation, and it is where I want to focus. **Selected Prior Experience** - **OpenClaw multi-agent orchestration (StreamIO AI):** Implemented gateway protocol, subagent delegation, profile management, and session switching — enabling coordinated AI agent workflows across four industry verticals. Directly analogous to designing and deploying Cresta AI agents across distinct customer use cases. - **RAG pipeline and multi-provider LLM orchestration (Fintellect AI):** Architected retrieval pipeline with ChromaDB, multi-provider LLM orchestration (Claude, GPT-4, Gemini) with fallback routing, structured output validation, and token budget optimization — the same class of infrastructure decisions required when building reliable AI agents for enterprise contact centers. - **ICE Self-Service platform (Intuit):** Delivered developer platform that reduced onboarding from 2–3 weeks to minutes in pre-prod and under 24 hours for production, while mitigating $1M+ in projected opex growth. Demonstrates ability to own a complex, cross-functional deployment lifecycle with measurable outcomes and senior stakeholder accountability. - **675M+ ICE engagements, 275% YoY growth (Intuit):** Scaled platform throughput from 6K to 50K TPS via rSocket migration supporting ~1.5M concurrent connections with sub-25ms TP99 across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma. Evidence of platform-scale thinking applied to production systems. - **aeval evaluation platform:** Built adversarial safety testing, refusal detection, data contamination detection, and statistical rigor (bootstrap CIs, effect size) into a CI/CD-integrated evaluation pipeline — the discipline needed to validate AI agent behavior before go-live and detect regressions post-deployment. - **Scheduler Service delivery at Splunk (4 months, end-to-end):** Owned a microservice from roadmap through production launch in approximately four months, demoed at .conf19. Demonstrates the bias for action and delivery urgency the JD calls out explicitly. - **Customer discovery and go-to-market execution (StreamIO AI, Fintellect AI):** Led customer discovery interviews, iterated product based on trader and user feedback, executed App Store launch, and established industry partnerships — directly applicable to facilitating workshops with customer stakeholders and aligning on agent use cases and success criteria. **Closing** Cresta's mission — making every conversation a competitive advantage — is one I find compelling precisely because it is measurable. AI agents either deflect calls, increase revenue, and improve CSAT, or they do not. I want to be accountable to those numbers, working at the boundary between customer strategy and technical execution. I would welcome the opportunity to discuss how my background in agent orchestration, platform-scale product delivery, and hands-on AI development maps to what Cresta is building. Thank you for your consideration. Sincerely, **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info