evan
banner
evanking.io
evan
@evanking.io
AI/ML researcher working with on-device audio, speech, and language models. occasionally exploring the intersection of code, ML, and design.

ece phd @ ut austin, 2024
cs bs @ unm, 2017

github.com/evmaki
If a tool increases productivity, people will adopt it organically and get a performance advantage. It’ll be so useful you can’t NOT use it. Forcing it from the top down seems to run the risk of actually reducing productivity if your assumptions about how useful it is are wrong.
November 15, 2025 at 11:43 PM
RTF is 10, so 50 sec to generate a 5 sec sample. Not exactly real time, but transformers takes 7 minutes to do the same. I was surprised to find there aren’t many people working with MusicGen models despite them being out since ‘23
November 13, 2025 at 7:12 PM
i'm assuming it's using some cloud-based model (LLM?) to do the filtering, which might raise privacy concerns. would be great to provide this locally, on-device.
May 22, 2025 at 11:28 PM
i’m going to have to go with the classic “Truck”. it’s simply goated in my opinion.
September 11, 2023 at 2:00 AM
including the middle name when he could have secured the OG firstnamelastname before i did…. Big time L from our friend ETK
September 8, 2023 at 3:50 AM