How it works
- 01
We map your work and size the machine
- 02
We buy the hardware for you
- 03
We set it up in your office
- 04
Your team goes live
Use Cases
Conflicts and intake agent
Desktop miniBilling guideline agent
AI workstationPrivilege review agent
tinyboxFirm knowledge agent
Blackwell server
Prior authorization agent
Desktop miniCharge capture agent
AI workstationAmbient scribe for every room
tinyboxDenial appeals agent
Blackwell server
School Medicaid billing agent
Desktop miniSpecial education compliance checker
AI workstationState reporting agent
tinyboxGrants and E-Rate agent
Blackwell server
Lead response agent
Desktop miniLease abstraction agent
AI workstationCAM reconciliation agent
tinyboxAcquisitions diligence agent
Blackwell server
Bookkeeping agent
Desktop miniTax document agent
AI workstationNotice response agent
tinyboxAudit agent
Blackwell server
The hardware
- Desktop miniMac mini or a small AI PCAbout $2,500Small models like Gemma 4 12B, gpt-oss-20b and Qwen3.6-27BA few people
- AI workstationNVIDIA DGX Spark or Mac Studio M5About $7,000 to $10,000Models up to about 120B, like gpt-oss-120b and Qwen3-Coder-Next10 to 25 people
- tinyboxtiny corp, up to 4 GPUsAbout $15,000 to $25,000Mid to large models, fast, with many requests running at once25 to 100 people
- Blackwell server8 NVIDIA RTX PRO 6000 GPUs, 768GB$100,000 and upThe largest open models, like DeepSeek, plus training on your own workHundreds of people
Maintenance
We hand it off running, with your team trained. Or we keep it current every month.
The Frontier
FAQ
What hardware do you like?
We like tiny corp's tinyboxes and machines built on NVIDIA or AMD GPUs, but we will source whatever hardware the job actually needs.
Who owns the hardware?
You do. The machine sits in your office and stays there after the install is finished.
Which models run on it?
Open models picked for each job, like Qwen, DeepSeek, Llama and Mistral. Small models handle simple work and larger ones handle the jobs that need them.
Can it use cloud models too?
Only if you want it to. By default, every request stays on your own hardware.
Do you come to our office?
Yes. We're based in New York and do every install in person, wherever your office is.
What happens after the install?
We hand it off with your workflows running and your team trained on them. If you want, we maintain it for you on a monthly plan: updates, new models and fixes.