Automations, APIs and LLMs in practice

In short

I'm a hands-on builder who connects business systems with automations and LLMs, and delivers solutions people actually use. I've spent 5+ years shipping production automation for B2B companies, across e-commerce, sales, marketing, field operations and healthcare-adjacent teams.

Make and n8n

I know Make and have built automations there, but n8n is where I'm strongest. It's my main orchestration layer: multi-stage workflows, webhooks, retries and validation loops, error alerts and AI agents with tools. The thinking transfers directly between the two (triggers, routers, iterators, error handling), so moving between them is a matter of syntax, not of approach.

Integrating APIs, Shopify in particular

I integrate with the APIs a business actually runs on. In my Shopify project I combined the Shopify Admin API (GraphQL), Klaviyo, Gorgias, Stripe and n8n into four automations: abandoned checkout recovery, replenishment reminders timed to each customer's own buying rhythm, an AI support agent with read-only access to live order data, and a payment reconciliation flow where Stripe webhooks mark Shopify orders as paid.

That last flow shows how I treat integrations that touch money: webhook signature verification, idempotency checks, exact-match validation, an audit trail on every update, and a Slack alert for anything that doesn't match instead of guessing.

Beyond Shopify, I've worked with Odoo, Gmail, Google Calendar, LinkedIn, Airtable, Webflow, Chatwoot, Publer and WhatsApp. At Adoreal I've built integrations with Vista Social (social media), Missive (email and newsletters) and ClickUp (task intake).

Working with LLMs

LLMs are a daily working tool and a component inside the products I ship. At Adoreal I work with Claude Code to design, build and test full-stack tools: a Partner Portal for practices (Next.js, Supabase, Vercel), and a Before & After vault that handles patient photos and consent under strict privacy rules (signed links only, real deletion, synthetic data outside production).

I focus on connecting LLMs to real business systems rather than demos. In production that has meant a WhatsApp agent for field technicians, a multi-agent SEO pipeline with RAG on Pinecone and Langfuse for observability, and Claude drafting LinkedIn comments that a person reviews before anything is published. I also built Groundwork, an app where business and tech teams agree on a project's brief, with an MCP server so agents can read the brief and propose changes without rewriting it.

Practical solutions across business areas

• Sales. CRM leads contacted and scheduled automatically: time to first contact went from 72+ hours to under 3 minutes and the meeting rate from 4% to about 13%.
• Marketing and content. Blog articles produced in under 10 minutes instead of 3 hours, and social carousels generated end to end with automatic visual validation, for multiple clinic brands.
• Operations. Field technicians generating service reports in minutes instead of days, straight from WhatsApp.
• E-commerce and support. Recovered checkouts, repeat purchases, instant order answers and reconciled payments for a Shopify brand selling in three markets.
• Compliance and internal tooling. Tools that handle patient photos and consent under strict privacy rules, and internal apps that keep business and tech aligned.

How I work

Across marketing, operations, healthcare-adjacent compliance and internal tooling, my approach is the same. I start from the real workflow, keep a human approval step where the stakes are high, put guardrails on data and costs, and test end to end before calling something done.

What I haven't done yet

I haven't worked with AWS yet. It's nothing that can't be achieved with the documentation and Claude.