I build production-grade AI-native software, context-engineering workflows, agentic developer tools, and product systems.

My work sits at the intersection of software architecture, product execution, and applied AI. I care about turning messy real-world codebases into systems that are understandable, operable, and safe for humans and agents to evolve.

• Context engineering for codebases — rules, memory, specs, hooks, review workflows, and agent operating layers for brownfield projects.
• AI-native product systems — LLM-backed workflows with queues, evaluation loops, observability, and product constraints.
• Full-stack product engineering — TypeScript, NestJS, React, Electron, Go, PostgreSQL, Redis, BullMQ, and cloud integrations.

• setup-context-engineering — a Claude Code Agent Skill that bootstraps real codebases with AGENTS.md, rules, prompts, memory docs, hooks, and a multi-agent PR reviewer based on the patterns it finds in the code.

Most of my recent work lives in private product repositories. I am turning the reusable parts into public architecture notes, labs, and playbooks:

• Electron + Go process orchestration — desktop app architecture, typed IPC, child-process lifecycle, and proxy health checks.
• YouTube signal intelligence — niche monitoring, reference libraries, AI-assisted packaging, and outcome measurement.
• WhatsApp commerce and booking agents — multi-tenant flows, intent parsing, checkout/appointment handoff, funnel analytics, and eval harnesses.
• Living feature knowledge — extracting and maintaining product knowledge from code, specs, meetings, and delivery workflows.

• I prefer small, verifiable systems over broad abstractions.
• I document decisions close to the code and keep operational memory explicit.
• I use agents as engineering operators, not as magic: they need context, boundaries, tests, and review.

• GitHub: @lucascruz18