SEODACH — an entire AI ops company, built solo

Context

An AI operating system built single-handedly to run a services business.

Problem

A small business needs lead-gen, CRM, content and operations without a large team.

My role

Sole architect & engineer.

Approach

Self-built MCP server (single entry, 5 scopes) over org-memory (Postgres + pgvector hybrid search: vector+BM25+trigram+RRF); 27 role-specialized agents on Claude/GPT through one LLM gateway with cost & limit control.

Outcome

27 agents and 32 microservices in production, automating ops end to end.

Stack