Zhanghao Wu

@Michaelvll1

co-founder & cto @ SkyPilot | AI Compute Platform for Frontier Teams

Joined March 2020

513

Following

2,210

Followers

Replies Reposts Media

Pinned
Zhanghao Wu
@Michaelvll1
Mar 19
Autoresearch from @karpathy runs 1 experiment at a time. We gave it 16 GPUs and let it run them in parallel. 8 hours. 910 experiments. 9× faster to the same best result. The most surprising part: the agent had access to both H100s and H200s. Without being told, it noticed H200s
SkyPilot
@skypilot_org
Mar 19
Karpathy's Autoresearch is bottlenecked by a single GPU. We removed the bottleneck. We gave the agent access to our K8s cluster with H100s and H200s and let it provision its own GPUs. Over 8 hours: • ~910 experiments instead of ~96 sequentially • Discovered that scaling model
21
Zhanghao Wu
@Michaelvll1
Sep 5
Can totally relate to this. It just gets into your life and stays. Love bufo, the frog
Megan O'Leary
@meganeoleary
Sep 4
in the software industry, bufo emojis are considered sacrosanct today, i investigated his origins as acting emoji ombudsman (self-appointed) this is his story
Article
A Portrait of Bufo as a Young Frog
Bufo entered my life on a February afternoon. We met over Slack, introduced by one of our wonderful customers. In a matter of days Bufo had permeated our internal channels. You see, he is a chameleon,...
1
Zhanghao Wu
@Michaelvll1
Sep 3
Congrats to the team on the launch! I worked with many of them during my time at MIT and always impressed by the talent density at @NunchuxAI. Great to see this multimodal inference service out! Looking forward to what comes next.
Nunchux AI
@NunchuxAI
Sep 3
Nunchux is live at nunchux.ai! We’re building the frontier of multimodal generative AI inference: fast, affordable, high-quality serving for image, video, and world models. Our Modelverse brings 30+ image, video, and avatar models behind one API. Use
Zhanghao Wu
@Michaelvll1
Sep 3
Congrats to the @inco_ai team!
Inco AI
@inco_ai
Sep 3
Inco AI is entering public beta with the ⚡️ fastest ⚡️ #inference endpoints for #Kimi K3, #MiniMax M3, #GLM 5.3, and #GLM 5.3 Flash. Ranking Number 1 on the respective @ArtificialAnlys provider leaderboard for output speed. Read more: inco.ai/blog/inco-plat…
Zhanghao Wu
@Michaelvll1
Sep 2
We have been using this for our day-to-day development of SkyPilot! Fully on our own compute / data with the GLM-5.3 model we serve internally. It manages a team of AI software engineers that do not share our proprietary code to external. Check it out on SkyPilot Platform.
SkyPilot
@skypilot_org
Sep 2
Introducing SkyPilot Agent Sessions. Start an agent, close your laptop. It keeps working on your own K8s cluster, and the dashboard tells you the moment it needs you: from the CLI, the browser, or your phone.
00:00