Production AI Systems

Multi-agent workflows, RAG pipelines, and automation tools. All self-taught, all running in production.

8 Agents Built
5 Running 24/7
10+ APIs Integrated
Python LangGraph LangChain ChromaDB Docker Claude API GPT-4o Gemini Whisper Google APIs ClickUp API Slack Bot SQLAlchemy Pydantic FFmpeg YOLO

Story Engine - Multi-Agent System

A complete video storytelling workflow built on my proprietary storytelling methodology.

Story Engine Production

Orchestrates 7 specialized AI agents through a LangGraph state machine with human approval checkpoints. Implements the "Ralph Loop" for automated quote validation ensuring 95% accuracy against source transcripts.

Agent Architecture

DirectorAgent

Creates story outlines using 6 plot points

ResearchAgent

Executes filtered RAG queries

WriterAgent

Generates scripts with VO and soundbites

Validator

Ralph Loop - ensures quote accuracy

XMLGenerator

FCPXML, EDL, markers export

MusicAgent

Music cues mapped to emotional arc

BRollAgent

Shot lists using 4 Layers of Place

16K+
Soundbites in RAG
95%
Quote Accuracy
7
Agents Orchestrated
3
Export Formats
LangGraph ChromaDB Claude API Whisper Pydantic Typer CLI FCPXML EDL

Media Toolkit Pipeline

Automated organization of 2TB+ footage with AI-powered analysis.

Media Processing Pipeline Production

Scans, categorizes, and catalogs large video/photo libraries with AI-powered scene analysis. Generates contact sheets, HTML catalogs, and searchable databases. Overnight batch processing with resume capability.

86,342
Files Scanned
40,256
Files Organized
2,242
Contact Sheets
34
HTML Catalogs

Pipeline Phases

Phase 1
Scan & Metadata
Extract duration, resolution, codec via FFprobe
Phase 2
AI Analysis
Claude Vision scene tagging, YOLO people detection
Phase 3
Organize
Move files to structured folders with rollback support
Phase 4
Thumbnails
Generate contact sheets with parallel processing
Phase 5
Catalog
Build searchable HTML catalogs with dark theme
Claude Vision YOLO FFmpeg SQLite Pydantic Python

24/7 Agent Suite

5 Docker agents running on NAS infrastructure.

Morning Briefing Production

Daily 6:30 AM briefings via Slack. Aggregates emails, calendar events, and news into prioritized summary using GPT-4o classification.

20
RSS Feeds
20
VIP Contacts
GPT-4o Gmail API RSS Slack
Calendar Assistant Production

Voice-enabled calendar assistant. Whisper transcription for voice notes, natural language scheduling, automatic event extraction from emails.

2
Calendars Synced
Voice Commands
Whisper GPT-4o Google Calendar Slack
Task Planner Production

ClickUp integration for AI-powered weekly planning. Natural language task queries, ICS calendar export, MCP server for Claude Code.

40
Hours/Week Planned
MCP
Server Built
ClickUp API Claude ICS MCP
Pics.io Router Production

Automated digital asset routing. Polls inbox every 5 minutes, routes based on metadata to teacher-specific collections.

28
Collections Mapped
5 min
Poll Interval
Pics.io API Docker REST
Notes Capture Production

Voice-to-note capture for personal knowledge base. Gemini transcription, smart folder routing, automatic tagging.

7
Folders Routed
Notes Captured
Gemini Slack Obsidian
Finance Analysis Development

Spending pattern analysis and visualization. Categorizes transactions and generates insights.

-
In Development
-
Coming Soon
Python Pandas

Production Deployment

All agents run on home server infrastructure with automatic recovery.

NAS
├── morning_briefing     # Daily briefings
├── martha               # Calendar management
├── crista_planning      # ClickUp integration
├── picsio-router        # Asset routing
└── obsidian_agent       # Note capture

Shared:
├── Google OAuth credentials (volume mount)
├── Automatic restart on failure
└── Cloud sync for code updates
                

Want Custom AI Tools?

I can build production-ready AI agents tailored to your creative workflow.