Can a small law office or clinic use AI on confidential client files without sending anything to the cloud? In this demo, three employees on three different devices chat with the same private knowledge base, and all of it runs on hardware in the room.
SETUP IN THIS DEMO
– RAMDeck pooling memory across a mini PC, a Mac mini, an old laptop and an RTX 3060
– A 30B model (about 17GB) split across the devices, running at roughly 38-39 tokens per second
– 13 fictional client documents for a made-up firm, "Harbor & Finch LLP"
– Three "employees": two MacBooks and an Android phone, all on the same local network
CHAPTERS
0:00 What you'll see: three devices, one private AI
0:24 Why confidential data needs a local AI
0:57 The 30B model setup and speed
1:25 Meet the fictional law firm
1:51 Three employees, one knowledge base
2:22 Uploading the client files
3:41 Checking the documents are spread across the nodes
4:00 Before and after: the model with and without context
5:31 Three devices, one shared knowledge base
6:09 Employees asking questions at the same time
7:51 Checking that the answers are accurate
8:26 Summary
NOTE: All people, clients, cases and documents shown are fictional demo data.
This video is not legal or compliance advice. If you handle regulated data, check your own requirements.
Early-access hub price: $109 (retail $199)
👉 https://www.indiegogo.com/en/projects/trademav/ramdeck-pool-your-ram-run-bigger-ai
LEARN MORE: https://ramdeck.com
More demos: https://www.youtube.com/playlist?list=PLBAnaNU1r3-M
RAMDeck — Pool Your RAM. Run Bigger AI.
Run and train models too big for any single machine, using the PCs, Macs, mini PCs and phones you already own.
Leave a Reply