Each keypress is appended to an 80 line prompt (key name along with timestamp of keypress and current text shown on the screen) and fed to a frontier LLM. Some of the office staff banged on the keypad for a few hours to generate training data to fine-tune the LLM on the task of denouncing key presses.
Thanks to some optimizations with Triton and running multi-GPU instances, latency is down to just a few seconds per digit entered.
You see, we needed to hit our genAI onboarding KPIs this quarter…
Na each key press goes to a separate lambda invocation that gets submitted to a Kafka queue, and what happens after that is a mystery to all involved.
We can make crazy latency ourselves just fine, no space transmission necessary
No, not a mystery, in fact.
Each keypress is appended to an 80 line prompt (key name along with timestamp of keypress and current text shown on the screen) and fed to a frontier LLM. Some of the office staff banged on the keypad for a few hours to generate training data to fine-tune the LLM on the task of denouncing key presses.
Thanks to some optimizations with Triton and running multi-GPU instances, latency is down to just a few seconds per digit entered.
You see, we needed to hit our genAI onboarding KPIs this quarter…
Probably a Celeron-powered PC tower barely keeping up with Windows Server 2008 R2 in a closet of a public office ;)
Gotta have multiple AZs.
The server is probably running Python
lol it's the flask debug server, "don't use this in production" banner and all