NVIDIA

Product Research Engineer, Deep Learning

AI Makerspace

GitHub Org: https://github.com/AI-Maker-Space

Co-founder & CTO

FourthBrain

Instructor and Curriculum Developer

Founding Machine Learning Engineer

Explorer Hop

Data Scientist

Turtle.AI

Worked on a team of 3 to build an end-to-end ZSL (zero-shot-learning) model for setting and detecting custom wake words!

"Turtle.AI" Wake Word Project

I am obsessed with Machine Learning, Generative AI, LLMs, and more — and I love sharing knowledge as much as I love learning!

For me, this is the biggest news of the year.

Colab in VS Code.

https://lnkd.in/g__aWQfh

Chris Alexiuk

Hey! 👋

My name is Chris, nice to meet you! 🙂

There's nothing I love more than Machine Learning and AI! 🖥️ 📖

Outside of my day job as a Founding MLE, you can catch me instructing and engaging with awesome people @ https://www.fourthbrain.ai/ 🧠

----------------------------------------------------------------------------------------------

I'm always down to chat about Machine Learning - Data Science - Statistics - Physics!

You can reach me at chris@alexiuk.ca or here on LinkedIn!

AI @ NVIDIA - 🏗️ Build | 🚢 Ship | 🚀 Share

for everyone worrying about AI taking over all the jobs: This is the CEO of one of the world's most advanced frontier model shops around, celebrating that their smartest model will stop using em-dashes when you ask it to stop.

Well, an incredible weekend wrapped at HackUTD (the largest 24hr. Hackathon in NA, btw) in Dallas. The next generation is, as the kids say, cracked.

We had over 100 submissions to our Nemotron track - so Shashank, Ethan, and I had the extremely difficult task of narrowing things down to the Top 3. In the end, we wound up with 3 projects that we felt best represented our open family of Agentic Models:

3rd Place - Data Driver: Mohammad Mehrab, Jackson Parrish, Jean-Pascal Gauthier, and Huy Nong. Who showed how Nemotron models could help automate data centre maintenance.

2nd Place - Alchemetrics: Ashrita Dara, Isha Rao, and Joanna Romesh for their project that used Nemotron to do live monitoring and decision making.

1st Place - Toyotron: Brice Duke, David Santos, Anthony Nguyen, and their final team mate G. Dinh, who built an incredibly featured auto-buying assistant application. It's no wonder they also won Toyota's track! 

Definitely hit any of these teams up if you're looking for well rounded builders!

Overall, the hack was dope - the organizers did an absolutely phenomenal job, HackUTD puts on the best show I've ever seen - no joke.

To everyone who spent the weekend building, win or not, you're the real ones - there are too many people and honourable mentions to....mention, but everyone we talked to has a bright future ahead.

Training a model family usually involves a massive trade-off: do you burn the compute to train Small, Medium, and Large variants separately, or do you settle for just one size?

The chefs at the NVIDIA Research kitchen just cooked up a fascinating solution to this bottleneck.

It’s called Nemotron Elastic, and it introduces a "Many-in-One" framework for training reasoning LLMs. Instead of training three separate models, the team trained a single 12B parent model that contains fully functional 9B and 6B submodels nested inside it.

The efficiency stats on this are pretty dope:

- 360x cost reduction compared to training a model family from scratch.  
- 7x cheaper than current SoTA compression techniques.  
- Zero-Shot Slicing: You can extract the smaller models instantly for deployment without any additional fine-tuning.
  
What’s really cool here is that they applied this to hybrid Mamba-Attention architectures. They solved the challenge of keeping the long-context capabilities required for reasoning tasks, even while compressing the model.  

Huge congrats to the research team for another banger, Ali Taghibakhshi and team killin' it as usual.

📄 Read the full paper here: ​https://lnkd.in/gGB8-_kW

Chris "The Wiz 🪄" Alexiuk

So stoked to be in Dallas for HackUTD! There are so many people here, and they are ready to build - the great weekend lock-in.

We're here to spread the good word of NVIDIA Nemotron - and see what awesome things people can build with the models in a weekend. Can't wait to see what the hackers cook up.

Also, I'm here with the incredible Shashank and Ethan, and we've been asking people what they're planning to build all morning - and let me tell you, it's gunna be good!

Had a blast today at the Toronto Zoo with my wife, during a much appreciated day off work; Shashank, Ethan, and I got a Hackathon to run this weekend in Dallas (UTD)! Hit me up if you're in the Dallas Area, btw!

While there, we learned about something called a "frozen zoo", essentially a large collection of genetic materials collected and maintained over many years that are useful in species conversation and, they're working on it, Jurassic Park style restoration. Mind blown, btw.

This is going to be an absolutely insane connection, but it really did remind me of recent work Anthropic is doing to "preserve Claude", via keeping the weights.

It makes you think about how humanity as a whole loves preservation, and the obsessive collection of things.

Really good day, really looking forward to building all weekend.

Just a reminder that NVIDIA has an open model for just about any task you could imagine.

Not just models, either. Datasets, code, technical reports, and more.

I've never been part of an organization that ships as much as the Team Green researchers and engineers.

https://lnkd.in/g2xbMXmB

Open article: nvidia (NVIDIA) by huggingface.co

Finally got around to setting up my recording area in my house.

Streams and videos 'boutta get leveled up!

Need a few more wall dressings for the wall behind me though, what should I put there?

"How do I get hired at XYZ org?"

There's 3 things I would prioritize above all else if you're looking to work in the AI space:

1. Work on your communication, no matter how good you think you are.

I learn something knew when it comes to communicating every single day. Either how to communicate with a new collaborator, or how to communicate my ideas better, or how to communicate more clearly, you get the point.

2. Please, for the love of God, learn your "handshakes".

Code interview? That's a handshake - if you know the handshake, you're more likely to get it. Random questions you feel are unrelated to the job? Another handshake, they show you can learn what you're expected to learn quickly, and thoroughly, which is what you'll need to do on the job. Behavioural interview? Believe it or not, that's a handshake. Don't put on an act, be who you'll be day in and day out, there's a company that is aligned with your work-style and vibe.

3. Understand the tech stack, and know how it fits into the ecosystem.

If you're applying to say, an AI software role at NVIDIA, and you don't know what Nemo is - it's just so much harder to explain and convince the hiring manager how you fit in, and how you're uniquely equipped to help. If you know the stack, and are passionate about some piece of it, you have a way easier time convincing someone that you belong doing that thing.

It's tough out there, with a load of talent, and AI tools complicating everything - but I truly believe that these things can help you stand out, and help you demonstrate why you're the right fit for whatever role you're after.

I know these things worked for me, and I hope they work for you as well. Stay sane out there!

Motivation for this post: I've gotten an influx of people asking me for a referral to NVIDIA recently, and while I don't refer people unless I'm familiar with their work - I do want to ensure that I give something of value to people who want to work at a tech company, and aren't sure where to get started or how to get noticed.

Google DeepMind  released Gemini 3 today, and I gotta say - I feel about this model the same way I feel about every Gemini model: "mid" (except on multimodal tasks, where it really shines bright - especially this time around.

The benchmarks are great, the hype train has definitely left the station and if chuggin' along full steam ahead - but to actually *use* the model leaves me a bit...whelmed.

I've tested it on a number of tasks and apps that I like to run new models through, and honestly it's just kind of another frontier model.

Also, it made this cute SVG of an otter riding a surfboard - and officially made Simon Willison bump the difficulty of his "pelican on a bicycle" test - so, it can't be that bad!

Once again. Hugging Face is crushing it at unprecedented levels with their newest book on training LLMs.

Honestly, there's something for everyone in here. From data selection to which optimizers to use to mid-training tips. This is one of the most *fun* ways to read about a highly complex topic.

Also, I'm obviously biased, but I loved seeing them use NVIDIA's Llama Nemotron post-training data. Truly epic.

This weekend, even if you're not likely to train a model - I'd recommend using AI to find parts that may interest you, and give this a read. It's well worth it.

https://lnkd.in/gHfhu37C

It's never a bad time to learn some CUDA, and we just dropped a new video series on exactly that!

This series will take you through a lot of the fun ins and outs of writing "proper" CUDA code (C++) and is a fun watch even if you don't want to take programming GPUs super seriously. You'll still learn a lot!

I love the idea that we're heading into a brave new world, when low level programming will be more important than ever, and these kinds of tutorials are great introductions to those fundamental patterns needed to make things go BRRRR on the hardware!

https://lnkd.in/gHUnpnvi

Happy Claude-mas to all that celebrate! Anthropic has done it again, and released yet another model that feels great to use - and crushes code tasks.

Opus 4.5 shows that we've nearly saturated SWE-bench verified (including in multilingual coding), and it's safer than ever. 

They've also added the "reasoning effort" feature we're seeing everywhere else (did Nemotron start the trend? No, but it was one of the first!) - letting you choose (through the API) if you want big think or small think!

Additionally, they did some updates to their apps - classic stuff. 

Overall - this model *feels* great to use, it is great at coding tasks, and it's not too slow for the performance. 

I'm happy to finally have a new daily driver model.

Shizhe Diao made this awesome fork of Nanochat that improves the model by a considerable amount, just by swapping out the data for some open Nemotron Data - for the same cost, as well!

The mix, leveraging both Nemotron-ClimbMix and Nemotron-Post-Training-Dataset-v2, leads to improvements through all of pre-, mid-, and post-training.

The basic idea here is simple: Good quality data is all you need!

Read more about what makes Nemotron-ClimbMix so well suited for pre-training small language models in the technical report, and check out the repo here if you want to produce an even better Nanochat.

Both links will be in the comments!

In Canada, November 11th marks Remembrance Day - a day to remember those who paid the ultimate price in War.

On this day, we wear Poppies over our hearts, and often we reflect on a poem, which I recently realized is a rather specifically Canadian poem, and I thought I'd share it here:

In Flanders fields the poppies blow
Between the crosses, row on row,
    That mark our place; and in the sky
    The larks, still bravely singing, fly
Scarce heard amid the guns below.

We are the Dead. Short days ago
We lived, felt dawn, saw sunset glow,
    Loved and were loved, and now we lie,
        In Flanders fields.

Take up our quarrel with the foe:
To you from failing hands we throw
    The torch; be yours to hold it high.
    If ye break faith with us who die
We shall not sleep, though poppies grow
        In Flanders fields.

Today was a big day! Ai2 released the newest OLMo model - which is absolutely openmaxxed. OLMo 3, and the accompanying.... everything, is well worth digging through. There are a lot of great tips for all stages of training, data curation, how to think about architecture, and more.

As well, Google DeepMind dropped Nano Banana Pro, which is basically just "in every way better" Nano Banana.

It's been a busy week, and the space shows no signs of slowing down.

Sometimes at my job, they just let me post stuff - Chains of Thought are powering the world, and it's not that hard to understand!

https://lnkd.in/gQnayhzg

Another great week to 🏗️ Build 🚢 Ship and 🚀 Share.

What are you building? I'll go first: A web applet for Draw Steel, a fun Tabletop RPG. I've been enjoying it, but there's some statuses and resources I'd like to be easier to keep track of!

Yessss. so many open models. so many retrieval models moved from closed to open

people are sleeping on this release, NVIDIA dropped a new family of Nemotron RAG models 🔥

it comes with text retrievers (sota!), multimodal retrievers as well as layout detectors with commercially permissive license 👏 
they even have Omni retrievers (image-text-audio!) 

find all the models here https://lnkd.in/d8P3wVWM

taplio

Chris "The Wiz 🪄" Alexiuk's Linkedin Analytics

Chris "The Wiz 🪄" Alexiuk

Performance ScoreChris "The Wiz 🪄" Alexiuk's LinkedIn Performance

Chris "The Wiz 🪄" Alexiuk's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Chris "The Wiz 🪄" Alexiuk's Linkedin Analytics

Chris "The Wiz 🪄" Alexiuk

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Chris "The Wiz 🪄" Alexiuk's LinkedIn Performance

Chris "The Wiz 🪄" Alexiuk's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Chris "The Wiz 🪄" Alexiuk's Linkedin Analytics

Performance ScoreChris "The Wiz 🪄" Alexiuk's LinkedIn Performance

Chris "The Wiz 🪄" Alexiuk's Best Posts (last 30 days)