
Karl Podesta
2.3K posts

Karl Podesta
@KarlPodesta
Irish techie. #HPC specialist @Azure in Europe. Forget data, #cloud brings #bigcompute to everyone! 20 years loving #HPC, #science, #opensource, #linux
Dublin City, Ireland Katılım Ocak 2014
3.7K Takip Edilen1.1K Takipçiler
Sabitlenmiş Tweet

A quick primer 🧵on #Microsoft #Azure #HPC for #SC24!
(Got pulled into meet, stuck in elevator, or beside one of us in the coffee queue? Just wanna know more? This might help!)
- Compute
- Storage
- Networking
- Software
- History
- Customers
- People
- Headline SC
English
Karl Podesta retweetledi

Kudos to Microsoft, they're helping to get OpenClaw ready for enterprises.
Omar Shahine@OmarShahine
New in @openclaw beta: one path scheme to rule them all. `openclaw path read|write|append` works the same across md, jsonc, jsonl, and yaml — plugins and agents now share a single addressing substrate for surgical edits to structured files. Nice work Gio Della-Libera — one of many PRs from the Microsoft Project Lobster maintainer crew! github.com/openclaw/openc…
English
Karl Podesta retweetledi

Read how we designed and deployed Multipath Reliable Connection (MRC), a multi-company step towards Ultra Ethernet, in Microsoft's datacenters.
Demonstrated with a 75k GPU pretraining job running stably through many faults.
📰 openai.com/index/mrc-supe…
📜 arxiv.org/abs/2605.04333
English
Karl Podesta retweetledi

At the Salishan HPC conference last week, I gave a talk called "AI doesn't need massive supercomputers after all!"
A couple people asked me for the slides (as crappy as they were), so here they are in hastily written blog format.
blog.glennklockwood.com/2026/05/ai-doe…
English

Time for me to buy that yacht I’ve had my eyes on.
That’s how these fundraises work, right?
Nicole Hemsoth Prickett@NicoleHemsoth
Such an exciting day here... bloomberg.com/news/articles/…
English
Karl Podesta retweetledi

We’re bringing our growing MAI model family to every developer in Foundry, including …
· MAI-Transcribe-1, most accurate transcription model in world across 25 languages
· MAI-Voice-1, natural, expressive speech generation
· MAI-Image-2, our most capable image model yet
Start building: microsoft.ai/news/today-wer…
GIF
English
Karl Podesta retweetledi
Karl Podesta retweetledi
Karl Podesta retweetledi
Karl Podesta retweetledi

Jeff Hammond from NVIDIA gives some historical context on NCCL's lineage going back to MPI, similarly for NVSHMEM. #AI is rooted in a strong #HPC networking history!
youtube.com/watch?v=2xMzQ1… (04:00)
Learn more about NCCL details in our paper: arxiv.org/abs/2507.04786

YouTube


English
Karl Podesta retweetledi
Karl Podesta retweetledi

blogs.microsoft.com/blog/2025/11/1… excited to continue to bring the widest set of advanced AI offerings to our customers.
English
Karl Podesta retweetledi

Microsoft Ignite 2025 Book of (Agentic) News: news.microsoft.com/ignite-2025-bo… #msignite
English
Karl Podesta retweetledi

Karl Podesta retweetledi

Microsoft's Ultra Ethernet 1.0 tutorial & workshop is now available on youtube 🎥!
Saurabh Dighe gives a brilliant motivation for Microsoft's "AI first datacenters" 🤖 followed by A.l Kabbani and myself on technical details and innovations of UE.
microsoft.com/en-us/research…




English
Karl Podesta retweetledi

Another first for our AI fleet... a supercomputing cluster of NVIDIA GB300s with 4600+ GPUs and featuring next gen InfiniBand.
First of many as we scale to hundreds of thousands of GB300s across our DCs, and rethink every layer of the stack across silicon, systems, and software to support next gen AI workloads.
English
Karl Podesta retweetledi
Karl Podesta retweetledi

My colleagues at NERSC just published a fantastic, comprehensive report on quantum computing and its implications for science. Surprised that >50% of their #HPC workload stands to benefit from #quantum computing. Excellent stuff.
osti.gov/biblio/2588210
English
Karl Podesta retweetledi
Karl Podesta retweetledi

🚀 Impressive ISCA’25 paper dissecting DeepSeek’s 🧠 compute stack: arxiv.org/abs/2505.09343
💡 FP8 precision, MLA, MoEs, Multi-plane Fat Trees w/ endpoint routing via NCCL’s PXN!
📐 The topology? Equivalent to a 1D HammingMesh—for ~2k nodes. 2D scales: arxiv.org/abs/2209.01346
English



