cover_letter / art_AwQl06Uck6w

Dear LangChain Hiring Team, LangChain sits at the center of one of the most consequential infrastructure shifts in software: moving AI agents from demos into production systems that enterprises can actually rely on. That mission resonates directly with work I have been doing for the past year — building multi-agent orchestration frameworks, RAG pipelines, and evaluation infrastructure from scratch as a founder, and before that scaling developer platforms to 675M+ annual engagements at Intuit. When I read the Deployed Architect job description, I recognized the exact intersection of infrastructure depth, agent engineering, and customer-facing technical leadership that I have been living. **Technical Foundation** My agent engineering work is grounded in production systems, not prototypes. At Streamio AI, I designed and implemented OpenClaw, a multi-agent orchestration framework built around a gateway protocol, subagent delegation, and profile/session management — coordinating AI agent workflows across real estate, insurance, health, and financial-markets verticals. That system runs on top of a FastAPI backend with 40+ agent tools, SSE streaming, enforced mutation approvals, and idempotent endpoints. For Fintellect AI, I architected a RAG retrieval pipeline using ChromaDB as the vector store, with multi-provider LLM orchestration across Claude, GPT-4, and Gemini, fallback routing, structured-output validation, and token-budget optimization — exactly the RAG patterns and knowledge organization work called out in the JD. On the evaluation side, I built aeval, a local-first model evaluation platform covering factuality, reasoning, instruction-following, safety, and code generation, with adversarial safety testing, refusal detection, and data contamination detection via SHA-256 hashing. Statistical rigor was a first-class concern: bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, and saturation detection. The stack — FastAPI orchestrator, TimescaleDB, Redis job queue, Next.js dashboard, Ollama — mirrors the observability and evaluation infrastructure LangSmith provides at scale. I also built a 3-phase RL post-training workbench benchmarking GRPO, DPO, PPO, and nine other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL, with GPU Docker passthrough and live SSE metric streaming on Apple Silicon and CUDA. My infrastructure background spans cloud deployment, CI/CD, and platform engineering. At Intuit, I led the ICE Self-Service platform (DevPortal, GitOps config, ICE Playground), reducing developer onboarding from 2–3 weeks to under 24 hours for production while mitigating $1M+ in projected opex growth. I scaled ICE throughput from 6K to 50K TPS via rSocket migration supporting approximately 1.5M concurrent connections at sub-25ms TP99. I led a Mailchimp GCP-to-AWS migration for MSaaS, delivering a Golang service template, MySQL persistence integration, and updated DevPortal documentation. I also built a Java JAR library for configuration drift detection across Git repositories as part of a broader MSaaS Drift Detection and Resolution program — the kind of GitOps hygiene work that maps directly to Terraform and Helm-based IaC practices. **Why This Role** The Deployed Architect role combines the two things I find most technically interesting right now: designing production-grade agent systems and working directly with enterprise customers to make those systems reliable. Having built and shipped AI products through the full lifecycle — from architecture through App Store approval and customer discovery — I understand both the engineering tradeoffs and the customer communication work that makes deployments succeed. What specifically draws me to this role is the scope: Kubernetes cluster design, multi-agent system architecture, evaluation framework design, and prompt optimization A/B testing, all in service of enterprise customers who are moving from PoV to production. The LangSmith platform — observability, evaluation, deployment, fleet management, and the newly launched Engine for autonomous agent improvement — is the infrastructure layer I would want to be building on. Working across LangChain, LangGraph, and Deep Agents with customers at the scale of Fortune 10 companies is the kind of technical challenge where my combination of platform engineering depth and hands-on agent development is directly applicable. **Selected Relevant Experience** - Implemented OpenClaw multi-agent orchestration framework (gateway protocol, subagent delegation, profile management, session switching) coordinating AI agent workflows across four industry verticals. - Architected RAG retrieval pipeline (ChromaDB vector store) with multi-provider LLM orchestration (Claude, GPT-4, Gemini), fallback routing, structured-output validation, and token-budget optimization. - Built aeval evaluation platform: 5 eval types, adversarial safety testing, bootstrap confidence intervals, Welch's t-test, Cohen's d, CI/CD regression detection, and automated safety gates — FastAPI, TimescaleDB, Redis, Ollama. - Scaled ICE platform to 675M+ engagements in FY23; led rSocket migration from 6K to 50K TPS supporting ~1.5M concurrent connections at sub-25ms TP99. - Delivered ICE Self-Service DevPortal with GitOps config, reducing developer onboarding from 2–3 weeks to under 24 hours for production. - Led Mailchimp GCP-to-AWS MSaaS migration; built Java drift-detection JAR scanning Git repos for configuration drift with OpenRewrite-based remediation roadmap. - Built production AI applications across FastAPI, Python, TypeScript/React, with Kinde OAuth, Apple IAP, Stripe, and multi-cloud deployment (EC2/nginx, GCP, AWS) — shipped signed/notarized cross-platform builds with full CI/CD and Playwright/Jest release gates. - NeurIPS 2014 published researcher (protein structure prediction); original 2004 system hand-coded in C++ with custom BPTT; 2026 rewrite spans 413 to 8B parameters across 5 neural architectures with MLflow, Optuna, and Docker orchestration. **Closing** LangChain's mission — making intelligent agents ubiquitous and helping developers move from prototypes to production — is the exact problem I have been working on from the builder side. I would bring to the Deployed Architect role a combination of infrastructure engineering depth, hands-on agent and evaluation platform development, and the customer-facing and communication experience that comes from founding and shipping products, teaching at De Anza College, and presenting at DeveloperWeek and Splunk .conf. I would welcome the opportunity to discuss how my background maps to what your Professional Services team is building. Sincerely, **O. Felix Amoruwa** famoruwa@berkeley.edu | 909-731-9011 | felixamoruwa.info