I build AI systems that takethe manual work out ofyour business.
- AI products in production, built end to end
- 2
- third-party bot cost replaced by one agent I built
- $12k / mo
- tax returns processed by the platform whose AI backend I own
- 8,000+
- CPA firms on that platform (platform figure)
- 119
How I think
I don't start with technology.I start with your problem.
Two products in production.Two built in the open.
Systems that do the workyour team does by hand.
- 01
AI agents for support and sales
Agents that answer customers, qualify leads and resolve tickets using your real business data, not a script.
Shipped: a sales agent that replaced a $12k/month chat bot at Commercient. - 02
Document intelligence
Read invoices, forms, contracts and scans in any condition, and turn them into structured data you can verify against the source.
Shipped: tax-document extraction for DTAXO across W-2s, 1099s, K-1s and brokerage PDFs. - 03
Workflow automation
Remove the manual steps between your systems one at a time, with retries, logging and human checkpoints where they matter.
Shipped: an AI prep and first-review pipeline that hands finished returns to RPA. - 04
Copilots and internal tools
Plain-language interfaces over your data. Ask a question, get a view, a report or an action, without waiting for someone to write SQL.
Shipped: a SQL view generator over synced ERP and CRM data. - 05
Self-hosted and fine-tuned models
Open-weight LLMs on your own servers when cost, privacy or control matter more than a per-token API bill.
Shipped: vLLM serving on in-house GPUs, with fine-tuned models for domain tasks. - 06
AI inside your existing product
Built into the software you already run, whether it is .NET, Python, your ERP or your CRM. Not bolted on beside it.
Shipped: agents integrated into a live .NET platform with paying customers.
A loop,not a list.
Start with what is actually costing you time or money, not with a tool.
One name.The whole system.
A direct line
WhatsApp, email or a call. You get answers from the engineer, not from a ticket queue.
End-to-end ownership
Discovery, architecture, models, backend, integration, deployment, and the fixes after launch.
Built inside your stack
I shipped AI into a .NET product with paying customers. I don't ask you to replace what already works.
Measured, not promised
Every project ends with numbers you can check: cost replaced, hours saved, tickets resolved.