EthosPowerAIDelivery: software built on your own data
I build automations, AI assistants, websites and web apps, connectors and complete platforms. They work from your own data, including data that customers share only with you. I prototype with your customer in weeks. The finished system is open source, can run entirely on your own servers and belongs to you. A person approves each decision, and no build replaces a member of staff.
Automations
Stop doing it by hand
Some work is moved by hand between systems that do not connect: test reports into the asset register, approvals and handovers, notifications, links to old systems. I automate it.
What this looks like in practice
A field engineer records battery discharge-test results by voice: voltage, temperature and discharge current for up to 54 battery blocks. An AI agent running on the company's own server captures them.
AI Agents
An assistant that knows your records
Assistants that research, draft, sort and answer from your own records, not the public web. They can support customer service, sales, research or multi-step workflows. I build them with LangChain or LangGraph, using Claude, OpenAI or a local model, and Qdrant or ChromaDB for search.
What this looks like in practice
An energy-sector company was losing leads because follow-up was manual. I built a system that checks the ERP and CRM for a matching email, company or phone number before it scores a lead. Duplicates merge into the original record. New contacts get three welcome emails, after a person approves them.
Websites & Web-Apps
A website, or a tool your team logs in to
Websites that answer an engineer's question from your own documents, and tools that your staff and customers log in to: dashboards, portals, intake and approval apps. This website is built the same way, on Next.js, Payload CMS and Cloudflare Workers. You own the code.
What this looks like in practice
ebatt.ai began as spreadsheets, became Django web apps and is now a live battery-quoting system. IEEE 485 sizing and EN 50272-2 venting are free to use. Product selection, quote requests and checkout follow them. One developer built it, so your first tool can start small too.
MCP Connectors
Let your customers' AI tools query your systems
A customer asks their AI before they ask you. A Model Context Protocol (MCP) connector lets that AI query your systems directly, so it gives your answer, not a guess. You set the access rules, and every query is logged.
This site runs one at /api/mcp, which serves its engineering knowledge base. Point your own assistant at it to see how a connector behaves before you order one.
Platforms
A complete system
This is the battery example from the approach page, built as one system. Charger test results go in. The system forecasts replacements, checks stock and lead times and schedules the site work. A person approves the purchase order. It runs on your servers as one system, not five separate tools. It starts with a one-week Discovery.
What this looks like in practice
An engineering consultancy wanted its people to search decades of reports and design documents with AI, without exposing them to cloud models or to anyone outside the firm. I built a knowledge base on the firm's own servers, with document management and end-to-end encryption. The archive never leaves those servers.
All prices in USD (converted from NZD), exclusive of GST. Each stage has a fixed price, agreed before it starts. The price ranges are on the Services page. I do not bill by the hour.
Not sure which of these fits?
Start with a call. If a small automation solves the problem, I will say so.