Samuel Rock
Engineering Manager · Hands-on · Self-taught
Started as a customer support agent. Learned to code on the side. Built a platform handling 5 000 support messages a day. Led a team from zero. Still shipping.
What I own end-to-end
Not a list of technologies — domains I've shipped in production and can reason about.
Ops Platforms
Omnichannel case management used by 70+ agents. 5K messages/day. SLA automation, multi-channel routing, homemade CRM on top of Zendesk API.
Engineering
Full-stack ownership from legacy PHP to TypeScript/Node/React. CI/CD, Docker, GitHub Actions, Prometheus/Grafana. 100% test coverage on modern stacks.
AI & Automation
Self-hosted n8n with custom nodes. Local LLMs via Ollama in production workflows. Claude agents for reporting, monitoring, and decision support.
Payments & 3PL
PayPal & Klarna dispute/claim workflows via API. RFN logistics: inbound/outbound, inventory, AdValorem insurance. Colissimo billing automation.
Partner Integrations
10+ production integrations, each with its own rate limits, failure modes, and retry logic. Shipped and maintained in production over several years.
Leadership
8 developers recruited. 3 PMs managed. Scoping gate, PR culture, post-prod monitoring. One team member grew into Head of User Care.
Projects
Open source work and production systems.
AgileFlow AI
Agile project management platform that bridges product vision and engineering execution, with configurable AI agents (Claude, GPT-4, Ollama, GCP).
claude-marketplace
LivePlugin marketplace for AI-assisted development — scope writing, architecture review, ADR management, deployment safety checks.
claude-config
LiveMy global AI workflow configuration — reusable agents, slash commands, and engineering conventions across all projects.
sr-ai-gateway
In progressUnified AI gateway with intelligent routing across Claude, OpenAI, Ollama, and GCP Vertex AI. Cost optimization + fallback logic.
sr-auth-service
In progressProduction-ready auth microservice — JWT access/refresh tokens, OAuth (Google, GitHub), MFA, audit logs.
How I work
From a vague idea to a shipped product — a repeatable method built for scale.
Define
User research, personas, user stories with acceptance criteria. Every feature starts with a documented problem — scope.md, ADRs, OKRs — before a single line of code is written.
Architect
Technical architecture with explicit decisions documented. Conscious trade-offs: monolith vs microservices, SQL vs NoSQL, build vs buy. All challenged by a dedicated @challenger agent.
Build with AI
Agentic development with Claude Code — specialized agents for analysis, implementation, and testing. TDD by default: tests written before code, every time.
Ship & Measure
CI/CD, automated migrations, health checks, Prometheus metrics. Deployment is boring by design. Success is validated against the OKRs defined in step 1.
Blog
Thoughts on engineering management, AI, and product development.
First articles coming soon.