jaberjaber23

6 ms·

31.
by jaberjaber23 1y ago
We profile and optimize kernels live on real GPUs!! So we’re different than unsloth
32.
by jaberjaber23 1y ago
Yup, that video shows our profiling reports:D
33.
by jaberjaber23 1y ago
Totally, NSight is great. We do something similar: generate kernels, profile them on real GPUs, then optimize based on that:D
34.
by jaberjaber23 1y ago
we’re adding those soon!!
35.
by jaberjaber23 1y ago
2x-4x improvements are normal when starting from a naive kernel, but sometimes we see gains well over 10x. Every kernel is profiled live on real GPUs (serverless), so you get accurate performance data for the specific architecture. Before-a
36.
by jaberjaber23 1y ago
totally agree. we're not trying to replace deep CUDA knowledge:) just wanted to skip the constant guess and check. every time we generate a kernel, we profile it on real GPUs (serverless) so you see how it runs on specific architecture
37.
We Made CUDA Optimization Suck Less(rightnowai.co)
47 points by jaberjaber23 1y ago|28 comments
38.
by jaberjaber23 1y ago
We’re RightNow AI. We built a tool that automatically profiles, detects bottlenecks, and generates optimized CUDA kernels using AI. If you’ve written CUDA before, you know how it goes. You spend hours tweaking memory access, digging through
39.
by jaberjaber23 1y ago
hey:D i built a tool that lets any dev write and optimize cuda code without knowing gpu stuff. just tell it what you want, it generates the kernel, runs it, and shows you bottlenecks. no setup, no cuda experience needed. we launched 3 days
40.
by jaberjaber23 1y ago
Hey everyone, I’m a solo founder and I launched RightNow AI ( https://www.rightnowai.co/ ) 3 days ago. It’s a tool for CUDA engineers (you write what you want, and it gives you optimized CUDA kernels + profiles them on server
41.
by jaberjaber23 1y ago
Hi HN, I am a solo founder building RightNow AI https://www.rightnowai.co/ , a tool that profiles, detects bottlenecks, and optimizes CUDA kernels using AI. The idea started from my own frustration with CUDA development. P
42.
Tell us why this is the worst product(rightnowai.co)
2 points by jaberjaber23 1y ago|1 comments
43.
by jaberjaber23 1y ago
I built an AI dev tool that helps you write and optimize CUDA code just by chatting. No deep GPU knowledge, no stress! Just vibe with the AI and (hopefully) get better performance. Roast it here: https://www.rightnowai.co
44.
RightNow AI v2 CUDA kernel optimizer with real-time profiling(rightnowai.co)
2 points by jaberjaber23 1y ago|1 comments
45.
by jaberjaber23 1y ago
After a ton of feedback, I've pivoted RightNow AI into v2: a browser-based chat interface that helps you write high-performance CUDA code without touching NSight or spending hours profiling. Just tell it what you're trying to do,
46.
by jaberjaber23 2y ago
Optimizing CUDA kernels is complex and usually requires expert knowledge. I built a tool that automates this process by profiling your code on serverless GPUs (matched to your GPU architecture), identifying bottlenecks, and applying SOTA op
47.
by jaberjaber23 2y ago
Optimizing CUDA kernels is often complex and requires expert knowledge. So, I built a tool that simplifies this process by profiling your code on serverless GPUs (matched to your GPU architecture), identifying bottlenecks in your kernel cod