Your data stays put
Models run on hardware you own or in your own cloud account. Nothing goes to a public AI service unless you decide it should.
Built like production
Monitoring, backups, access control and a runbook come standard. This is infrastructure, not a demo that breaks next month.
Ready for your rules
Texas Ethics Opinion 705, the FTC Safeguards Rule and Texas' AI law (TRAIGA) all put the duty on the firm. You get a written policy and a setup documented against the NIST AI Risk Management Framework.
Four steps, each at a fixed price you know before we start.
You can stop after any step. Most firms start with the first and decide the rest once they have seen results on their own files.
AI Safe Start
In about two weeks your firm has a list of the AI tools your staff already use, a written AI use policy ready for your attorney's review, business-grade tools set up with the right privacy settings, a trained team, and a ranked list of the three tasks most worth automating.
from $2,500 · fixed priceProof on your files
We run a sample of your own documents overnight on our private hardware and show you what came out, next to the originals.
included in Safe StartOne build
Pick one: private search across your files with citations, intake to first draft, or a document backlog turned into clean data. It runs on your server or your cloud account.
from $5,000Monthly plan
We keep it running and add one scheduled agent for a recurring job. Its work waits in an approval inbox, so nothing is sent without your click.
from $500/moAlso available
A private AI server in your office, fine-tuning on your own documents, DevOps and cloud retainers, and training for your team.
ask for a quoteWhat happens in the two weeks.
Inventory
We list the AI tools your staff already use and the client data each one touches.
Policy
A written AI use policy for your firm: approved tools, what may go into them, who checks the output. Your attorney reviews it.
Setup and training
Business-grade tools configured so your data is not used for training, and one session that shows your team how to use them.
Your next three
A ranked list of the three tasks most worth automating, with what each would save.
See it work on your own documents before you buy a build.
Send a sample set. We run it overnight on GPUs we own and operate ourselves, here in Texas, and show you the results next to the originals. No OpenAI, no hyperscaler, no third party in the middle.
- GPU memory
- 224 GB
- Cards
- 2× RTX PRO 6000 Blackwell · RTX 5090
- Runs
- open-weight models up to ~120B parameters on-site
- Location
- Texas, owner-operated
- Each client gets a separate, encrypted workspace, deleted when the proof is done, with written confirmation.
- What you depend on every day then runs on a server in your office.
- Prefer a cloud? We deploy into your own AWS, Azure or GCP account instead.
Run by an engineer who runs this at home.
HomeRack is run by Ahmer, a DevOps and cloud engineer with 14 years of production infrastructure. Ahmer's own lab runs a fleet of local and cloud models, agents that hand off work to one another, and a shared knowledge base they keep current.
Every client system gets the same standards the lab runs on: measured before it ships, monitored after, and written down.
- Experience
- 14 yrs DevOps & cloud
- Models
- local + cloud, open-weight
- Based in
- Texas