Gaurav Sahu

The entire industry is racing to make AI cheaper. Most people assume the answer is better chips or bigger GPU clusters. However, while the industry is buying its next 2× when there's a 1.78× speedup is sitting in the software.

We've been hard at work @unifiedsciences, building an autonomous optimization engine toward radically more efficient compute.

Today I'm very happy to share some early results where our unified engine:
1) wrote faster GEMM kernels for cuBLAS (up to 1.78x faster)
2) made an efficient AI model 6x smaller and 5x faster without any quality loss
3) optimized diverse everyday AI workloads on a consumer GPU!

this is cool!!

@AirCanada there's a major issue in your system: my meal preference doesn't make it to the staff in flight. my booking confirmation shows it, my boarding pass shows it, yet, the staff quite conveniently said, "so, there's no [my preference] meal available". i'm sure you can figure out the rest of the details.

fixing vibecoded ai slop is something i am very good at, so if you need help fixing the former, i'm willing to help.

also, where's the warmth and hospitality gone? you'll have to figure this one out on your own.

A fine research opportunity with a fine researcher—you’ll enjoy your time, highly recommend!

2 new terms to add to my vocab. nice piece by the @COLM_conf PCs!