Portfolio
AI Agents & Pipelines

Autonomous Social Media AI Pipeline

Multi-tenant content generation, design & publishing platform for LinkedIn, Instagram, Threads and Telegram — powered by microservice AI agents, RAG fact-grounding, and Kubernetes scale-to-zero.

GitHub Repository
20s
Content Generation
End-to-end from prompt to publish-ready post
90%
Cost Reduction
vs. traditional content agency pricing
Scale-to-Zero
Cloud Infra
KEDA event-driven autoscaling on K3s + Azure
~0%
Hallucination Rate
RAG-grounded facts + numeric validation
The Problem

Creating verified B2B content for LinkedIn, Telegram, and Instagram traditionally requires a full agency team — copywriter, domain expert, graphic designer, editor, SMM manager — costing $3,000–$10,000+/month with a 2–3 day cycle per post.

The Solution

A microservice AI graph orchestrated by OpenClaw via BullMQ/Redis: an AI writer produces copy grounded in a RAG fact database, a Puppeteer carousel renderer creates 1080×1350 branded images, a Telegram bot delivers a mobile HITL moderation cockpit, and a self-correction evaluator re-generates posts scoring below 85% alignment.

Architecture Highlights

01

Multi-tenant architecture with isolated brand style guides, color palettes & knowledge bases per tenant

02

Puppeteer carousel engine rendering 1080×1350 PNG branded graphics per post

03

Telegram mobile moderation cockpit — approve, edit, regenerate in 1 tap

04

Self-correction evaluator agent — auto-regenerates posts scoring <85% against golden datasets

05

KEDA Redis-triggered scale-to-zero: resource-heavy pods spin down when idle

06

High-speed LPU inference via Groq (llama-3.3-70b) + Gemini 2.0 Flash

Tech Stack

Node.js 20TypeScriptNext.js 15BullMQ / RedisKubernetes K3sKEDAPuppeteerGroq LLaMA 3.3Google GeminiAzureDockerPostgreSQL

Want something similar?

Available for Founding Engineer roles & architecture contracts.

Discuss ProjectAll Projects