John Zedlewski

This is a great #chartcrime example! Price over time, not adjusted for inflation. There was 35% inflation between 2016 and 2025. Uses a narrow time window and a clipped y-axis. Here is a longer series in 1984 inflation adjusted dollars: fred.stlouisfed.org/graph/fredgr...

YouTube video by NVIDIA Developer cuML Accelerates Machine Learning by 50x with Zero Code Change

You can train your scikit-learn #ML models up to 50x faster on GPU now, without changing a single line of code. The cuML accelerator for scikit-learn is now in open (and open source) beta): www.youtube.com/watch?v=cIJs...

Dude, you got to get it together and do some damn lorem ipsum today. Looks like you’ve been putting it off for AGES.

Nvidia boss saves California College of the Arts with mega donation The Nvidia boss swoops in as the school faces a budget shortfall, layoffs, and declining enrollment.

Nvidia boss saves California College of the Arts with mega donation

Data scientists and engineers, did you know that @pola.rs goes even faster if you have a GPU? NVIDIA and the Polars team are doing a webinar with a deep dive into how it all works on Jan 28. Check it out! info.nvidia.com/nvidia-polar...

I can personally confirm this! I am an NVIDIA employee with cancer, and from Jensen to the most junior member of my team, the company has been incredibly supportive, I am so lucky to be here.

Best Practices for Multi-GPU Data Analysis Using RAPIDS with Dask | NVIDIA Technical Blog As we move towards a more dense computing infrastructure, with more compute, more GPUs, accelerated networking, and so forth—multi-gpu training and analysis grows in popularity. We need tools and also...

Sometimes it's great to write a super-deep tech blog, but I also appreciate these shorter ones. Like, here we just want to give a quick rundown of exactly how to configure Dask so you can scale to multiple GPUs with the best possible memory/networking config: developer.nvidia.com/blog/best-pr...

Made-in-Berkeley induction stove from Copper lets users 'plug and play' The range operates on 110 volts so that users don't need an electrical upgrade, and a battery stores the power.

Maybe this bsky account needs to broaden to focus on GPUs, data science, and induction stove fandom? Seriously, induction is the best way to cook, and it's cool to see a company rolling it out with 110v power: www.berkeleyside.org/2024/11/21/m...

NVIDIA Releases cuPyNumeric, Enabling Scientists to Harness GPU Acceleration at Cluster Scale Researchers can use the NVIDIA cuPyNumeric accelerated computing library to effortlessly run their data-crunching Python code.

cuPyNumeric is a super cool approach - full NumPy compatibility in Python, but automatically scaling up to 1000-node HPC levels and supporting both GPUs and CPUs. You can try it now: blogs.nvidia.com/blog/cupynum...

NetworkX Introduces Zero Code Change Acceleration Using NVIDIA cuGraph | NVIDIA Technical Blog NetworkX accelerated by NVIDIA cuGraph is a newly released backend co-developed with the NetworkX team. NVIDIA cuGraph provides GPU acceleration for popular graph algorithms such as PageRank, Louvain…

Anyone working in #graph analytics knows that NetworkX is super popular for prototyping, but it can get quite slow. This blog introduces zero-code-change GPU acceleration for NetworkX code with massive speedups. (Think 100x or more for large graphs.): developer.nvidia.com/blog/network...

I'll be trying out some occasional posts here on BlueSky, mostly talking about #GPUs, #machinelearning, and interesting stuff we're doing at NVIDIA around accelerating data processing and data science.