Writing

Things I've shipped & learned.

Tutorials, framework comparisons, and case studies from 2.5+ years of building production AI and GTM automation systems.

Self-hosting n8n on GCP: A production setup with Docker, PostgreSQL, and Nginx

Self-hosting n8n on GCP: A production setup with Docker, PostgreSQL, and Nginx

I replaced $500+/month in SaaS automation subscriptions with a self-hosted n8n instance on GCP. Here's the exact Docker + PostgreSQL + Nginx + SSL setup, plus the 3 mistakes I made so you don't have to.

LangGraph vs CrewAI vs AutoGen: When to use each one — my honest comparison after shipping in all three

Comparison13 min

LangGraph vs CrewAI vs AutoGen: When to use each one — my honest comparison after shipping in all three

I've shipped production multi-agent systems in all three frameworks. LangGraph is the production-grade choice, CrewAI wins for fast prototyping, and AutoGen is best for research-style conversations. Here's the honest breakdown — with code, costs, and failure modes.

How I replaced a $2,000/month data subscription with a custom Clay pipeline that's actually fresher

Case Study9 min

How I replaced a $2,000/month data subscription with a custom Clay pipeline that's actually fresher

The client was spending $2k/mo on Apollo + ZoomInfo and getting 60-90 day stale data for their niche. We built a custom Clay pipeline with Google Maps scraping + LLM enrichment + HubSpot sync. Lead freshness went from 60-90 days to 7-14 days, SDR research time dropped from 3-4 hours to under 30 minutes per day.

Agentic AI in GTM & RevOps: What MCP Servers Actually Change in 2026

Agentic AI in GTM & RevOps: What MCP Servers Actually Change in 2026

MCP doesn't replace n8n or Zapier — it replaces the per-tool integration code underneath an agent. Here's where that matters for RevOps and where it's just this quarter's buzzword.

Claude Sonnet 5 vs GPT-5.6: Which Model Should Power Your AI Agents in 2026

Comparison10 min

Claude Sonnet 5 vs GPT-5.6: Which Model Should Power Your AI Agents in 2026

For most production agents doing multi-step tool calling, Claude Sonnet 5 is the default I reach for; GPT-5.6 Sol earns its keep on a narrower set of frontier-reasoning tasks. Here's the breakdown by workload, not by benchmark score.

The Best RAG Stack for B2B Knowledge Assistants in 2026 (Qdrant vs Pinecone vs FAISS vs AstraDB)

Comparison10 min

The Best RAG Stack for B2B Knowledge Assistants in 2026 (Qdrant vs Pinecone vs FAISS vs AstraDB)

Everyone obsesses over which vector database to pick. In production, that's usually the fourth-most-important decision you'll make on a RAG system.

AEO vs GEO vs Traditional SEO: What Actually Changed for AI Search in 2026

AEO vs GEO vs Traditional SEO: What Actually Changed for AI Search in 2026

Everyone is selling 'GEO' as a replacement for SEO. It isn't. Here's the actual mechanical difference between optimizing for Google's AI Overviews and optimizing for ChatGPT, Perplexity, and Claude — with the implementation I run on my own site.

How to Build an MCP-Powered Slack Agent for RevOps (Step-by-Step)

How to Build an MCP-Powered Slack Agent for RevOps (Step-by-Step)

I built an MCP-powered agent that checks HubSpot for stalled deals and posts a daily summary to Slack — here's the exact server, agent loop, and scheduling setup, mistakes included.

Want this for your team?

I build production AI systems — RAG, agents, GTM automation — for B2B teams worldwide. Book a 20-minute intro call.

Book a 20-min intro call