Agentic AI Builder | AI Operator | LATAM (Costa Rica)
I build systems that make operations faster and more reliable. My work focuses on moving beyond basic LLM wrappers to build production-ready architecture: multi-agent pipelines, webhook automation hubs, document intelligence agents, and rigorous evaluation frameworks.
Current focus: Agentic AI development using Claude Code, Python, and Trigger.dev.
I believe in building AI systems with genuine production signals: error handling, structured outputs, observability, and human-in-the-loop (HITL) design.
| Repository | What it demonstrates |
|---|---|
| claude-ops-agent | A triage agent with a strict Human-in-the-Loop (HITL) approval gate and structured observability logging. |
| maker-checker-pipeline | Two-agent quality control pattern with scored evaluation, feedback loops, and capped revision cycles. |
| doc-intelligence-agent | Document ingestion and chunking that extracts and validates typed JSON against a strict schema. |
| webhook-automation-hub | A Python-based webhook listener with Claude-powered classification and routing across business event types. |
| agent-eval-harness | A deterministic evaluation framework with test cases, latency tracking, category scoring, and run history. |
- Agentic: Claude API, Claude Code, Cursor, Trigger.dev, n8n, Zapier
- Languages: Python, JavaScript (Working proficiency)
- Integration: REST APIs, Webhooks, HubSpot, GoHighLevel
- Practices: Prompt engineering, Human-in-the-Loop (HITL) guardrails, structured JSON outputs, agent evaluation
Anthropic AI Fluency | DeepLearning.AI Agentic AI | AMP AI Operator | Google PMP
📫 Connect with me: LinkedIn | ninaverse.blog