What AI work have you shipped?

I built the AI pipeline for Pinnacle Drop, a SaaS that turns a supplier product link into store images and complete Meta ads with copy, sold on monthly AI credits with Stripe subscriptions.

How much does an AI feature cost?

One AI feature in your existing app starts at $99 (3 days). A document chatbot with sources is $349 (6 days). An AI product layer with admin uploads, per-user limits or credits, Stripe metering and conversation logs is $749 (12 days).

What will the AI cost me to run?

You get an estimate per 1,000 uses before I build, from tests with your data. Usage limits keep the bill predictable.

All Services

AI Features & Chatbots

AI that holds up with real users, at a cost you can predict.

A feature that writes, summarises or generates inside your app, or a chatbot that answers from your own documents and shows its sources.

Book a Free Call View Projects

From $99Cost estimate firstAnswers with sources

Why this service exists

Adding an API call to OpenAI takes an afternoon. Making it reliable takes more: answers that come from your documents instead of made-up facts, limits so one user cannot run up your bill, and a clear idea of what it will cost each month before you launch.

My process, step by step

The job and real examples

You show me the app and the job the AI should do, with 3 to 5 real examples of input and the output you want.

Tests on your data and a cost estimate

I test prompts on your real examples and send the results with an estimated cost per 1,000 uses, before I build anything.

Build with limits

I build it into your app with usage limits and error handling, on a preview link. For chatbots: ingestion, embeddings in Supabase pgvector and a chat UI with sources.

Handover

You learn how to change the prompt and watch usage and costs. Your API key stays in your hosting environment variables, never in the code.

Every engagement includes

OpenAI, Gemini or Claude in your app

Document chatbots with sources (RAG)

Usage limits, credits and Stripe metering

Cost estimate from your own data first

Common questions

Specific answers for this service engagement.

Yes. I build chatbots on the OpenAI API (or Gemini or Claude on request) that answer from your own documents, using embeddings in Supabase pgvector. Answers show their sources, and when nothing matches, the bot says so instead of guessing.