Gennady Pekhimenko

CEO and Co-Founder at CentML: Professor at the University of Toronto

[Soumith Chintala @soumithchintala]

Sometimes we forget that NVIDIA wins because it's a software company. DGX Spark is a reminder of that. It's a CUDA dev machine that's beautiful enough and small enough to be on my desk and with enough memory to fit a truckload of params.
It's not the fastest or best at anything, but it's great to develop on and transfer your final training run to a H/B200, final robotics policy to your Jetson, final inference to {nvidia/apple/amd/[favorite vendor]}.

[Andrej Karpathy @karpathy]

My sleep scores during recent travel were in the 90s. Now back in SF I am consistently back down to 70s, 80s.

I am increasingly convinced that this is due to traffic noise from a nearby road/intersection where I live - every ~10min, a car, truck, bus, or motorcycle with a very loud engine passes by (some are 10X louder than others). In the later less deep stages of sleep, it is much easier to wake and then much harder to go back to sleep.

More generally I think noise pollution (esp early hours) come at a huge societal cost that is not correctly accounted for. E.g. I wouldn't be too surprised if a single motorcycle riding through a neighborhood at 6am creates millions of dollars in damages in the form of hundreds - thousands of people who are more groggy, more moody, less creative, less energetic for the whole day, and more sick in the long term (cardiovascular, metabolic, cognitive). And I think that many people, like me, might not be aware that this happening for a long time because 1) they don't measure their sleep carefully, and 2) your brain isn't fully conscious when waking and isn't able to make a lasting note / association in that state. I really wish future versions of Whoop (or Oura or etc.) would explicitly track and correlate noise to sleep, and raise this to the population.

It's not just traffic, e.g. in SF, as a I recently found out, it is ok by law to begin arbitrarily loud road work or construction starting 7am. Same for leaf blowers and a number of other ways of getting up to 100dB.

I ran a few Deep Research sessions and a number of studies that have tried to isolate noise and show depressing outcomes for cohorts of people who sleep in noisy environments, with increased risk across all of mental health (e.g. depression, bipolar disorders, Alzheimer's incidence) but also a lot more broadly, e.g. cardiovascular disease, diabetes.

Anyway, it took me a while to notice and after (unsuccessfully) trying a number of mitigations I am moving somewhere quiet. But from what I've seen this is a major public health issue with little awareness and with incorrect accounting by the government.

[Darian Shirazi @darian314]

We're less than a month away from the @GradientVC AI Founder Summit in NYC on 6/5 during @Techweek_! This year we have an incredible lineup of AI experts and founders:
- @dwarkesh_sp fireside chat with @soumithchintala from @AIatMeta, founder of @PyTorch
- @zachweinberg from @Curie_Bio @flatironhealth
- @Dr_GGP from @CentML_Inc, @calcsam from @mastra
...and a secret unannounced speaker!

[Gennady Pekhimenko @Dr_GGP]

I totally agree. I told my students to prioritize publishing at systems conferences, MLSys, and release papers on arxiv quickly while simultaneously promote the open-source artifacts for their papers. I stopped reviewing for NeurIPS myself as the quality of submissions I got was very low.

[Gennady Pekhimenko @Dr_GGP]

I look at the concerns around #AI art the same way I look at coding.

Sure AI can do the work but we’ll always want a human involved to be part of the process.

The same can be said about AI art - you can’t teach an AI “taste”.

You need a human to determine what is “good” art.

At the end of the day AI is just generating images.

Humans are determining the art.

#ai #art #AIart

[Gennady Pekhimenko @Dr_GGP]

2025 product strategy:

✅ build the app everyone talks to

✅ then build the app everyone talks on

#chatgpt #openai #Building

[Gennady Pekhimenko @Dr_GGP]

Great results by @CentML_Inc and other top vendors in supporting #Llama4 day 1.

[Artificial Analysis @ArtificialAnlys]

Congratulations to @togethercompute, @FireworksAI_HQ , @databricks, @DeepInfra, @CentML_Inc and @GroqInc on having day-one Llama 4 inference endpoints live!

Keep an eye out for endpoints coming this week from @Azure, @cerebras, @SambaNovaAI and more.

Both @Meta's Llama 4 Scout and Maverick only have 17B active parameters - so although their total sizes are relatively large at 109B and 402B respectively, these models have the potential to enable extremely fast and efficient inference.

All providers serving Llama 4 so far are offering it at cheaper prices than their Llama 3.3 70B endpoints, including for Llama 4 Maverick. This makes Maverick as an incredibly compelling model with a wide range of inference options.

[Gennady Pekhimenko @Dr_GGP]

Amazing new collection of powerful @Meta #Llama4 models in partnership with @AIatMeta supported right away on CentML's Platform: https://t.co/asaa3jx2a4.

#Llama4 #meta #centml

[Gennady Pekhimenko @Dr_GGP]

AI isn’t just another productivity tool - it’s your Cybernetic Teammate

At least according to this Harvard study 👉 https://t.co/t9FjLIMWuz

✅ Solo employees using AI matched up to the performance of their human team counterparts

✅ Teams broke down silos using AI to create balanced solutions

✅ Those working with AI had positive social experiences over teams that didn’t use AI

✅ Less experienced employees levelled up using AI to the same standard as experienced employees

✅ AI enabled teams were 3x more likely to generate top 10% solutions

Seems like the most common hire companies will start making is that of the cybernetic kind.

#ai #teams

[Gennady Pekhimenko @Dr_GGP]

Before AI agents can think for themselves

They have to work their way up as digital middle managers governed by bureaucratic processes

like the rest of us.

[Gennady Pekhimenko @Dr_GGP]

🌲 Lumber – tariffed.
🛢️ Oil – tariffed.
🎓 AI talent – shaping the world for decades, still can’t be tariffed.

#ai #talent #Canada

[Gennady Pekhimenko @Dr_GGP]

What is the typical drama with America’s Next Top Model?

@perplexity_ai : Models often face personal issues or insecurities, resulting in tearful moments.

We’ve officially reached #AGI!

#topmodel #genai

[Gennady Pekhimenko @Dr_GGP]

@eric_haibin_lin Great work @eric_haibin_lin! Excited to see it presented at MLSys!

[Gennady Pekhimenko @Dr_GGP]

The average age of a Fortune 500 exec is ~57 yrs old.

The employees most likely to use AI to help them work are younger than 50.

Something tells me the lag in AI adoption in the enterprise may not just be about the tech...

[Gennady Pekhimenko @Dr_GGP]

Claude 3.7 Sonnet just pulled us out of the coding stone age.

#Claude37Sonnet

[Gennady Pekhimenko @Dr_GGP]

Everyone is talking about how @deepseek_ai used less resources and built an optimized model on par with industry leaders.

Now we should be talking about how we build optimized datacenters.

The software exists to make this happen.

For those investing in AI, less is becoming more.

Image from Reuters: European data centre space shortage expected in 2025 as AI booms (https://t.co/7ke6G5cnO7).
#ai #datacentre #datacenter #software #optimization

[Gennady Pekhimenko @Dr_GGP]

Moore’s law: the number of transistors in a dense integrated circuit doubles approximately every two years

@nvidia : Hold my beer

[Gennady Pekhimenko @Dr_GGP]

Currently trying to sort out my conference schedule for the year ...
#Llama

[Gennady Pekhimenko @Dr_GGP]

@Burchoff @neuralmagic @mgoin_ @CentML_Inc At @CentML_Inc we actually provided some of those chips on our platform: https://t.co/MyxFCCppZy

[Gennady Pekhimenko @Dr_GGP]

@AnthropicAI doesn’t want job applicants using AI.

An AI company doesn’t want people to use AI - it sounds like an SNL skit, but I get it.

Technology advances because of creative minds, not because of how well a model is trained.

When I’m hiring at @CentML_Inc I’m looking for YOU, not your prompt engineering skills.

And yes, we have a bunch of open roles we’re looking to fill, check them out below 👇

(And no, we’re not hiring for a prompt engineer)

https://t.co/fTfejOcx0b