
founder, builder, gen-zer | Head of Growth @vast_ai
Los Angeles, CA
Joined July 2011
- Tweets2K
- Following194
- Followers1.1K
- Likes9.7K
Sleeper GPU of the Week 💤
A new weekly series highlighting the most undervalued GPU on Vast.
First up: GB200. 744 GB of GPU memory across 4 GPUs. DeepSeek V4.1 Flash benchmarked at 3,000+ output tokens/sec across 64 concurrent requests. The whole server: $35/hr.
Proud to be among @brexHQ’s 25 fastest-growing software vendors this fall. Thanks to the builders choosing Vast and the GPU hosts making it possible. 🚀
Win Feigle retweeted
Do you use @vast_ai to spawn GPU instances from a large marketplace of available machines?
If you do, you can now mount HF Storage Buckets to your instances and unlock super fast/scalable storage, optimized for AI 🔥🔥
Great collab, give it a try its super convenient 🥰
We just keep growing.
⚡️ Vast.ai is back on Ramp's Top SaaS Vendors trending list for May — #6, up from #9 in March.
Ramp tracks real spend data across 50,000+ businesses. Vast.ai landed in the Trending category — breakout growth relative to size — the same signal that flagged Anthropic and Perplexity before they became household names.
Read more here: ramp.com/leading-indicators/…
@tryramp
We love using @tryramp and it looks like Ramp customers love using Vast.ai 👀
@vast_ai is a top 10 trending vendor among Ramp customers.
2025 @vast_ai Product Launch Event, live-streamed from SF at 7:00 PM PT tonight: youtube.com/live/rE9anL5AoNA
Win Feigle retweeted
🚨 Something big is coming from vast.ai next week.
We’re hosting a live product launch with an in-person event for investors and AI founders and you’re invited.
December 9, 2025. San Francisco + Livestream.
We’ll be unveiling the next evolution of Vast.ai.
𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿 𝗵𝗲𝗿𝗲:
→ Event in San Francisco: luma.com/atnnffe2
→ Livestream: youtube.com/live/rE9anL5AoNA
Tech prediction: the next breakout companies are going to be the ones that make AI feel invisible, not impressive.
The best AI teams stopped fighting their infrastructure months ago.
They figured out that manual provisioning kills momentum. 20 minutes to spin up GPUs means time lost to queuing instead of iterating.
Smart orchestration and automated provisioning turn GPUs into API resources. Teams ship faster by focusing on models, not infrastructure.
What would you build if provisioning wasn't the bottleneck?

















