AI Applications · Agents · Automation · Generative AI

I build AI systems thatactually do things.

AI agents, automation, generative media pipelines and custom applications — built, integrated and deployed.

Explore my work View GitHub

SYSTEMS ONLINE — BASED IN AUSTRALIA (AEST)

152+
AI agent skills
multi-agent orchestration
$200–550
/mo YouTube revenue
1 monetised channel of 4
100%
Self-hosted AI infra
VPS · local generation · n8n
6
Systems built
agents · automation · generative media
Featured systems

Six real systems. Full case studies.

Each one shows the actual architecture, what was built, the technology, and the honest status — including what's still in development.

ORCHESTRATORHERMESCODEpython · gitRESEARCHdeep · webBROWSERplaywrightMEDIAimage · video152+ SKILLS · 16 CATEGORIES · MCP TOOL ACCESS
01LIVE

Hermes

Multi-Agent AI Development Environment

Problem
Building increasingly complex AI workflows requires multiple specialised capabilities rather than one general-purpose chatbot.
Built
A local autonomous AI environment with 152+ specialised skills, multi-agent orchestration, research, coding, browser automation, media generation and business automation capabilities.
Result / Capability
Runs daily as the operating layer for every other project on this site.
PythonLLMsMCPPlaywrightAPIs
View case study
RESEARCHSCRIPTVOICEGENERATEimg + vidEDIT / CUTTHUMBNAILPUBLISH$200–550 / MONTH4 CHANNELS · 1 MONETISED · AI-ASSISTED PIPELINE
02LIVE

AI Content Production System

End-to-end faceless YouTube pipeline

Problem
Producing consistent YouTube content manually takes 10–20 hours per video across research, scripting, voiceover, visuals and editing.
Built
A full pipeline — research → scripting → AI voice → generated visuals → programmatic edit → thumbnail → publish — operating 4 channels, 1 monetised.
Result / Capability
$200–550/month demonstrated revenue from the monetised channel.
ComfyUITTSFFmpegHyperframesAgents
View case study
LOCAL GPU · RTX 5070 TiComfyUI · llama.cpp · TTSNO API CALLS · NO PER-TOKEN COSTIMAGE GENVIDEO GENVOICE · PT DE FR JA NLMUSIC GEN
03LIVE

Local Generative AI Lab

Image, video, voice & music generation on local hardware

Problem
Cloud generative APIs impose per-token costs, rate limits and privacy trade-offs that make high-volume experimentation impractical.
Built
A local-first lab on an RTX 5070 Ti: ComfyUI image/video workflows, local LLMs via llama.cpp, multilingual TTS (PT/DE/FR/JA/NL/EN-AU) and music generation — all offline.
Result / Capability
Unlimited local generation at zero marginal cost; feeds the content system.
ComfyUIllama.cppGGUFFFmpegPython
View case study
VPS · SELF-HOSTEDWEBHOOK INn8nWORKFLOWLLMPROCESSBASEROW DBCALENDARNOTIFY
04LIVE

AI Automation / n8n Infrastructure

Self-hosted workflow engine + data + integrations

Problem
AI models alone don't run businesses — they need to be wired into calendars, databases, notifications and external APIs reliably.
Built
A self-hosted VPS stack running n8n workflow automation, Baserow (data), Dograh (voice) with webhook/API integrations connecting LLMs to real business processes.
Result / Capability
The integration backbone every AI system on this site runs through.
n8nBaserowWebhooksDockerLinux VPS
View case study
DATA INjson · mdCOMPOSITIONcode-drivenFFmpeg · Remotion · Hyperframes · OpenMontageVERSION-CONTROLLED COMPOSITIONS · DATA → RENDER → PUBLISH
05COMPLETED

AI Video Generation Pipeline

Programmatic, code-driven video production

Problem
Manual video editing doesn't scale — 50 videos from a template means 50 manual edits, and nothing is reproducible.
Built
Programmatic video engines where compositions are code: data in → composition → rendered output, using Remotion, Hyperframes, OpenMontage and FFmpeg pipelines.
Result / Capability
Version-controlled, reproducible video output powering the content channels.
RemotionHyperframesOpenMontageFFmpegTypeScript
View case study
CALLER AUDIOSTTLLM AGENTINTENT + REPLYTTSTOOLSCALENDAR · CRMPROTOTYPE · IN DEV
06PROTOTYPE / IN DEVELOPMENT

AI Voice Receptionist

Voice agent for inbound business calls

Problem
Small businesses miss calls during work hours; every missed call is a lost lead.
Built
Prototype pipeline: caller audio → STT → LLM agent (intent + reply) → n8n tool actions (calendar/CRM) → TTS back to caller. Built on Dograh + n8n + local LLMs.
Result / Capability
Working prototype — call flow demonstrated end-to-end. Not yet production-hardened.
Dograhn8nLLMSTT/TTSWebhooks
View case study
Browse all projects by category
Open for work

Have something that needs building?

Remote roles, contract projects, or collaboration. I answer every serious enquiry — usually within a day.

Get in touch David_plos@hotmail.com