ThreeZinc AI Platform
Multi-Agent Orchestration System
40%
fewer pipeline failures
<200ms
p95 report target
100%
agent nodes traced
Overview
A multi-agent platform orchestrating Instagram and LinkedIn connectors for automated content generation and campaign analytics.
Owned the architecture of an AI platform that orchestrates multiple intelligent agents across Instagram and LinkedIn connectors for automated content generation and analytics, built on LangGraph-style agent graphs. Backend connector design consolidates fragmented platform data into a single centralised pipeline, with explicit handling for scraping edge cases, rate limits and schema drift — the three things that quietly break every social-data pipeline. Every agent node is instrumented with LangSmith tracing, and failure modes plus recovery strategies are documented in the repo rather than living in someone's head.
Stack
- LangGraph
- LangSmith
- Node.js
- TypeScript
- PostgreSQL
- Redis
- OpenAI
- Year
- 2026
- Focus
- AI/ML, Multi-Agent, Backend
- Role
- Architecture & delivery
More work
All projectsMulti-modal Generation Platform
Image, Video & 3D Inference Backend
A unified backend fronting image, video and 3D generative models, with async queues for compute-heavy inference.
- FastAPI
- Python
- Redis
- +3
