Professional workAI automation2025 — now

LLM Automation Pipelines

n8n + LLM pipelines and AI-enabled apps delivered as project lead at Freedom With AI.

01Why I built it

Client operations — notifications, reminders, planning, and task management — were handled through repetitive manual work.

Client operations ran on the same repeated work every day — sending notifications, chasing reminders, planning and updating tasks by hand. Work that repeats that predictably shouldn't need a person, so I set out to take it off the team's plate.

Questions I had to answer
  • Q1

    Which steps are pure rules, and which need the judgment an LLM can supply?

  • Q2

    How do we automate without turning the process into a black box the team can't change?

  • Q3

    How do we ship to production without interrupting client operations?

02Approach

Custom n8n + LLM automation pipelines replaced those workflows end-to-end, backed by Node.js and Python services and AI features in Flutter apps.

03Key decisions
  1. n8n as the backbone

    Visual pipelines keep every automation readable and editable, with LLM steps only where a decision or a written message is needed.

  2. Streaming backends in Node.js and Python

    REST APIs over PostgreSQL with LLM streaming and prompt orchestration, shared by the pipelines and the Flutter apps.

  3. CI/CD on AWS for every release

    Releases go out through pipelines with zero downtime, so automations people depend on keep running while we deploy.

04Architecture
streamClient workflowevents · schedulesn8npipelineLLM stepprompt orchestrationActionsnotify · plan · tasksFlutter appschat · voice · doc Q&ANode / PythonREST · streamingPostgreSQLon AWS
AIServiceDataInterfaceData flow
05AI components
01
LLM automation

n8n pipelines with LLM steps automate notifications, reminders, planning, and task management.

02
Prompt orchestration

Backends integrate LLM streaming and prompt orchestration.

03
Embedded AI features

Flutter apps ship chat, voice, and document Q&A connected to backend services.

06Implementation
  • Led delivery end-to-end: requirements, architecture, team coordination, production release.
  • Node.js and Python backends with REST APIs and PostgreSQL.
  • Production systems on AWS with CI/CD pipelines for zero-downtime releases.
07Tech stack
n8nLLMsNode.jsPythonPostgreSQLFlutterAWSCI/CD
08Looking back
The hard part

Leading it, not just building it — turning scattered requirements into an architecture, coordinating the team, and owning the release when real client work runs on it.

What I learned

Automation is judged by what people stop doing. The best pipeline is the one the team forgets exists, because the work simply gets done.

09Result

Manual client workflows replaced end-to-end, eliminating repetitive operations.