cover_letter / art_1lXQpPmxR5Y

Dear Cresta Hiring Team, Cresta is doing something genuinely consequential: taking the contact center — historically one of the most friction-laden surfaces in business — and rebuilding it around AI agents that can actually resolve problems rather than route them. The opportunity to own AI Agent outcomes at the customer level, from pre-sale scoping through post-go-live optimization, is exactly the kind of end-to-end product accountability I have sought out throughout my career. My path from hand-coding backpropagation in C++ at UC Berkeley to building production multi-agent orchestration systems today has been defined by one consistent thread: closing the gap between what AI can do in theory and what it actually delivers in a customer's hands. ## Technical and AI Foundation My hands-on AI work spans the full stack of what Cresta's platform requires. At Streamio AI, I designed and 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 was not a wrapper around an existing framework; it was a ground-up architecture decision about how agents hand off context, how delegation boundaries are enforced, and how session state is preserved across agent transitions. That is directly analogous to the design and build work Cresta's Forward Deployed PM role requires. On the evaluation side, I built aeval — a local-first model evaluation platform with five core eval types (factuality, reasoning, instruction-following, safety, code generation), adversarial safety testing with refusal detection, and statistical rigor including bootstrap confidence intervals, Welch's t-test, and Cohen's d effect size. This matters for a role that owns agent outcomes: knowing whether an agent is actually performing requires more than eyeballing transcripts. I also built an RL post-training workbench benchmarking GRPO, DPO, PPO, and nine additional algorithms across TRL, VeRL, OpenRLHF, and NeMo RL — giving me a working understanding of how the underlying models that power conversational agents are trained and optimized. My research foundation includes a NeurIPS 2014 accepted paper on neural networks for protein secondary structure prediction, and the original system behind that paper was a neural network I hand-coded in C++ with custom BPTT in 2004 — the 2026 rewrite scales from that 413-parameter original to 8B parameters. ## Why This Role The Forward Deployed PM role at Cresta sits at the intersection of customer-facing leadership and technical depth — a combination I have operated in repeatedly, and one that most PM roles do not actually require simultaneously. What draws me specifically to this role is the ownership model: you own the AI Agent and the outcomes it drives, full stop. That is the right accountability structure for this problem, and it matches how I have approached product leadership throughout my career. ## Role-Specific Connection Cresta's platform spans conversational AI agents, real-time human augmentation, and conversation intelligence — and the Forward Deployed PM is the connective tissue between all three, as experienced by the customer. I am particularly interested in the workshop and design session component of this role: facilitating sessions with senior customer stakeholders to align on agent use cases, workflows, and success criteria is a skill I developed at Intuit, where I conducted enterprise-wide assessments and presented strategic recommendations directly to the CTO. The feedback loop back into Cresta's platform roadmap — improving playbooks, surfacing new capabilities — is also where I can add disproportionate value given my background building developer-facing platforms at scale. ## Selected Prior Experience - **OpenClaw multi-agent orchestration (Streamio AI):** Built gateway protocol, subagent delegation, profile management, and session switching — enabling coordinated AI agent workflows across multiple industry verticals; directly applicable to designing and iterating on Cresta's AI Agent configurations with FDEs. - **ICE Self-Service platform (Intuit):** Delivered DevPortal, GitOps config, and ICE Playground, reducing developer onboarding from 2–3 weeks to minutes in pre-prod and under 24 hours for production — demonstrating the ability to own a complex deployment lifecycle end-to-end and measure outcomes in concrete time-to-value terms. - **275% YoY ICE engagement growth (Intuit):** Scaled platform to 675M+ engagements in FY23 across QuickBooks, TurboTax, Mint, Mailchimp, and Credit Karma; scaled throughput from 6K to 50K TPS via rSocket migration supporting ~1.5M concurrent connections with sub-25ms TP99 — evidence of operating at enterprise scale with measurable infrastructure and business outcomes. - **RAG retrieval pipeline (Fintellect AI):** Architected ChromaDB vector store, multi-provider LLM orchestration (Claude, GPT-4, Gemini) with fallback routing, structured output validation, and token budget optimization — the same architectural patterns underlying robust conversational AI agents. - **aeval evaluation platform:** Built adversarial safety testing, refusal detection, and CI/CD regression gates — giving me the tooling mindset to define and track the metrics that determine whether an AI Agent is actually working for a customer. - **Scheduler Service delivery at Splunk (4 months, June–Oct 2019):** Took a net-new microservice from zero to demoed at .conf19 in roughly four months — demonstrating the bias for action and delivery urgency the JD explicitly calls for. - **Enterprise stakeholder leadership (Intuit/Splunk):** Presented language investment strategy to Intuit CTO; led benchmark testing initiative for Fortune 500 beta customer at Splunk achieving up to 10x query performance improvements — comfortable owning relationships at the senior executive level and translating technical findings into business decisions. ## Closing Cresta's mission — turning every customer conversation into a competitive advantage — is one I find genuinely compelling, because the problem is hard in exactly the right ways: it requires deep AI capability, real deployment discipline, and the ability to earn trust from enterprise customers who have seen too many AI pilots fail to deliver. I have spent the last several years building the technical foundation and the customer-facing judgment to do all three. I would welcome the opportunity to discuss how my background maps to what Cresta is building. Thank you for your consideration. Sincerely, **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info