AI Agent Pipeline Development
Autonomous agent systems, multi-LLM orchestration, prompt engineering, and council debate architectures. Agents that run daily, not just respond to chat.
Trust on the record
Paul Sully — Production, Systems & Creative
Osaka, Japan — 37 years in Japan, 3+ years engineering production systems
Building production software. From autonomous AI agent pipelines to Web3 smart contracts, multilingual infrastructure to automated video production. Systems that run, not demos that impress.
Services
Each service line is backed by production systems and a network of specialist reviewers for human-in-the-loop quality.
Autonomous agent systems, multi-LLM orchestration, prompt engineering, and council debate architectures. Agents that run daily, not just respond to chat.
Smart contracts, wallet integrations, token economics, and deep due diligence investigations. Technical audits and structured evidence-backed reports.
Node.js, TypeScript, React, Python. APIs, databases, and front-end interfaces built for reliability and maintainability, not resume padding.
AI translation and generation, verified by native Japanese speakers. We hire human reviewers to check machine output, flag errors, and provide comparison data for training and quality improvement. Human judgment where it matters.
HyperFrames pipelines, voice-over generation, subtitle and caption systems, automated composition and rendering. Full pipeline from script to publishable video.
VPS provisioning, nginx reverse proxy, Docker container orchestration, continuous monitoring, and backup automation. Multi-server architectures that stay up.
Deep-dive investigations, evidence collection, cross-referenced source material, and structured reports with clear sourcing and uncertainty handling.
Portfolio
Each project represents completed, operating work under the Attest Dojo brand.
Trust Attestation Platform
Complete AI practice: website, branding, pricing model, and operating methodology. Multi-model orchestration, source-verified research, monitoring pipelines, and autonomous briefing generation.
Production system handling live due diligence, cross-referenced research, and monitoring. Structured for client engagements from $2,500 dossiers to $20,000 custom builds. Relevant for AI workflow design, agent orchestration, due diligence, research automation.
Agent Framework
Fleet of 20+ specialized AI agents running on Ollama, OpenRouter, and Claude. Council debate architecture (3-5 models debate decisions before execution), autonomous research loops, continuous improvement cycles, and 100+ LaunchAgents managing the fleet.
Runs daily — autonomous research briefings, security audits, content generation, and pipeline automation. The system that powers Attest Dojo's backend.
Multilingual Infrastructure
Bilingual content pipeline: daily briefings, video scripts, documentation, and client deliverables in Japanese and English. Cultural adaptation, Japanese typography, honorific handling, and business-formal Japanese.
Production-grade JP/EN content across web, video captions, email, and technical documentation.
Media Automation
End-to-end AI-assisted video production: scriptwriting via LLM, voice synthesis, HyperFrames HTML5 animation, automated bilingual captioning (JP/EN), and CapCut final assembly. Zero manual caption editing.
Multiple video series produced. Full pipeline from idea to publishable video.
Sample Video
AI-generated loop video produced entirely through our automated pipeline. Script, voice, animation, and compositing handled by agents with human review.
Automated research system monitoring news, regulatory filings, and market data across JP/EN sources. Daily briefings with source verification, cross-referencing, and structured summaries. Web archiving pipeline for evidence preservation.
Runs autonomously. Feeds into dossiers, monitoring reports, and due diligence workflows.
Infrastructure Portfolio
The physical and virtual infrastructure that keeps everything operational. Designed and maintained in-house.
Multi-Server Architecture
Mac Mini + cloud VPS + 4TB NAS. Local inference on the Mac, public-facing services on the VPS, bulk storage and backup on the NAS. Each machine has a clear role.
Ollama Local LLM Server
Multiple models running locally with GPU acceleration via MLX and Metal. Serves AI agents, council debates, and research pipelines without cloud API costs for routine work.
Automated Monitoring
Daily health check pipelines across all servers. Uptime monitoring, disk alerts, service status dashboards, and automated recovery scripts. Catches problems before they become incidents.
Nginx Reverse Proxy
Multiple local domains routed through a single nginx instance with SSL termination, caching, and access control. Public-facing services isolated from internal infrastructure.
Docker Orchestration
Containerized services for development, staging, and production. Isolated environments per project with consistent build pipelines and zero-downtime deployments.
Continuous Backup System
rsync-based automated backups with retention policies. Code, configs, databases, and media backed up on schedule. Tested restore procedures, not just backup scripts that might work.
Contact
If the work fits, the next step is a direct conversation, not a sales funnel. Reach out with a real project, a technical challenge, or a system that needs to run better.