Put AI to work inside your business
Custom AI systems and LLM integration for small businesses and startups
I build AI systems that take the repetitive work off your plate - replies, bookings, documents, reports - connected to the tools you already use. Get hours back every week.
I build AI systems that run inside your business - Claude connected to the tools you already use, workflows that run every day, assistants that know your data. Your team operates at 2x without switching tabs.
Velo - your AI workspace as a solo operator
A chat-first AI workspace for solo operators and fractional consultants juggling multiple clients. Built by me, now in private beta.
Join the Velo waitlistPrefer I build it for you? Book a free audit call
20 tools. 1 brain. Every dark corner of your data - visible.
From one-person businesses to startups
Same engineering, sized to you. You don't need a technical team - or a big budget - to work with me.
An assistant that drafts replies from your inbox, WhatsApp or website chat. Invoices and documents read and filed automatically. A weekly report from Gmail, Drive, Shopify, HubSpot or your accounting software - without anyone building it by hand.
GitHub, Jira, Slack, your databases, analytics and billing - connected to one AI your team works with every day. Answers across every system, routines that run on their own.
What this looks like for a small business
One annoying task, handled. No big project, no tech team needed.
The task: client emails and WhatsApp messages pile up, and every reply starts from zero.
What I build: an assistant that knows your services and prices and drafts each reply for you to approve and send.
The task: bookings, cancellations and reminders handled by hand every day.
What I build: reminders sent automatically, cancellations flagged, and a short daily summary of who's coming.
The task: the same customer questions about shipping and returns, over and over.
What I build: answers drafted from your own policies and order data, plus a weekly sales report from Shopify.
The task: invoices, receipts and documents arriving by email that someone has to read and file.
What I build: documents read, sorted into the right Drive folder, and key numbers sent to your bookkeeper.
No need to know what's technically possible. Describe your business and the task you'd love to stop doing - I'll reply with how I'd approach it.
Running solo? Try Velo, my AI workspace for solo operators (private beta).
What Claude can actually do in your org
Not demos. Not prototypes. Built and running in production.
"Why did retention drop last Tuesday?" Claude queries MongoDB, Postgres, and Mixpanel in one session and gives you an answer - no SQL, no analyst, no waiting.
Six months of decisions, ideas, and open questions sitting in #product and #engineering. Claude reads it, extracts the signal, and hands it back as a structured brief.
Claude reads your Elasticsearch logs, groups recurring failures, flags cost spikes, and tells you what's worth fixing - before your users notice.
Mixpanel funnel + Stripe MRR + MongoDB user data - pulled together in a single Claude session. No BI tool. No SQL. Just ask what you want to see.
Set it once. Every morning: yesterday's errors summarized and posted to Slack. Every Monday: sprint brief from Jira pulled and ready. Every Friday: business pulse from Stripe + Mixpanel delivered to #product.
Runs as Claude scheduled tasks - ready when your team opens their laptop.
What I connect
- GitHub
- Jira
- Confluence
- GitHub Actions
- MongoDB
- PostgreSQL
- Mixpanel
- AWS
- Slack
- Google Drive
- Gmail
- Figma
- Miro
- Stripe
- HubSpot
- Intercom
How it works
We map your current tools, data and daily routines, and find where AI saves you the most time. Then you get a fixed-price proposal sized to your business.
Build the AI system and connect it to your tools via MCP. Write the skills it runs on. Test against real workflows.
Your team works with Claude as the center. I stay available for the first month.
Priced to your business, not a package
A one-person shop and a 50-person startup need different things. So every project starts with a free conversation, and you get a fixed price before anything starts.
- Free audit call. We look at your tools, your routines and where your time goes.
- Written proposal. What I'd build, how long it takes, and one fixed price. No hourly meter.
- Build and hand over. I build it, connect it, show you how to use it, and stay available afterwards.
Writing
Technical posts on AI systems, LLM architecture, and engineering operations.
Questions people ask
Who is this for?
Solo operators and small businesses first: consultants, freelancers, clinics, studios and small shops that want AI doing real work - answering customer messages, processing documents and invoices, and producing reports from the tools they already use. And startups that want Claude connected to their actual data and workflow, not just chat.
Do you build AI systems for small businesses?
Yes. I build custom AI systems for small businesses: assistants that draft replies from your inbox, WhatsApp or website chat; document and invoice processing; and scheduled reports pulled from tools like Gmail, Google Drive, Shopify, HubSpot or your accounting software. No technical team needed on your side.
How much does it cost?
It depends on the size of the business and what needs connecting. Every project starts with a free audit call, after which you get a written proposal with one fixed price - no hourly meter. Small businesses and solo operators can start with a single workflow.
How long does a project take?
It depends on what is being built. A single workflow for a small business is a short project; connecting a whole startup stack takes longer. The written proposal after the free audit call states the timeline and one fixed price, and I stay available after handover.
What is MCP integration consulting?
I configure Model Context Protocol (MCP) connections between Claude and your engineering tools - GitHub, Jira, Slack, AWS, MongoDB, Stripe, and more. The result is one AI brain with access to all your systems, replacing context switching with a single interface.
What tools can you connect via MCP?
GitHub, Jira, Confluence, MongoDB, Postgres, Mixpanel, AWS, Slack, Google Drive, Gmail, Figma, Miro, Stripe, HubSpot, Intercom, and more. If there is an MCP server for it, I can wire it in.
Can Claude access GitHub, Jira, and Slack at the same time?
Yes. MCP allows multiple servers to run simultaneously in one Claude session. You can query GitHub for open PRs, check the linked Jira ticket, and post a Slack summary — all in a single conversation. That cross-tool capability is the core of the AI Operating Layer I build.
How do I connect Claude to Jira via MCP?
You install a Jira MCP server, configure it with your Atlassian credentials, and point Claude Desktop to it via the MCP config file. Claude can then read tickets, update statuses, and create issues in natural language. I handle the full setup and make sure it works reliably with your existing Jira workflows.
What is an AI operating layer for a startup?
An AI operating layer is when your AI (Claude) has live read/write access to every tool your team uses — GitHub, Jira, Slack, your database, your analytics. Instead of switching between apps, your team asks Claude and gets answers that cross all your systems. It replaces fragmented AI chat with a single intelligent interface wired into your actual work.