Skip to main content
Amazon SageMaker

Amazon SageMaker

Search complete. 20 mentions across 11 episodes found for "Amazon SageMaker".

Nov 17, 2026

Ravi ShankarGUEST
25:54
So we make the AI data ready and also from a context perspective.
Ravi ShankarGUEST
26:01
And with AWS, we are embedded within the core technologies for creating the AI agents, for example, on the bedrock, on the agent core side, within the ML models with the SageMaker and so on.
Ravi ShankarGUEST
26:15
So this actually enables us from a technology perspective, integrating them the data and the AI control pains together.
Ravi ShankarGUEST
26:24
And the second one is in terms of go to market, enabling their sellers to be able to be aware of what Denodo can provide and articulate that in the sales cycles for Denodo to be able to take that and educate our customers in terms of what AWS brings to the table.
Flo MotlikHOST
3:44
Plus a web search API for Agent Core, OpenAI's Daybreak cyber models on the Bedrock runtime endpoint, and OAuth 2.0 discovery URLs in agent core identity.
Flo MotlikHOST
3:56
SageMaker Jumpstart adds DeepSeq V4 Pro, V4 Flash, and V4.1 Flash, each with a 1 million token context, plus deep reinforces ornith agentic coding models.
Flo MotlikHOST
4:09
SageMaker Unified Studio adds designation-based layouts that control which tools each project member sees.
Flo MotlikHOST
4:15
Quick, the AI workspace, is renamed AWS Quick, and QuickSight becomes AWS QuickSight.
Flo MotlikHOST
4:22
Quick admins can now restrict which structured data source types users may create, update, or share.
Megha SrivastavaGUEST
1:49
And regardless of which option, they would then be wanting to use a much larger amount of compute resources and likely be turning to services that provide model training services for the client.
Megha SrivastavaGUEST
2:02
So this could be, for example, the OpenAI fine tuning API or Together.ai that also provides model training services or Amazon SageMaker.
Megha SrivastavaGUEST
2:10
So increasingly users might shift to a third party that has the compute resources needed to train these powerful models and give them their data set and ask those service to perform model training for them and then provide the model back to the user.
Megha SrivastavaGUEST
2:30
And so the problem that arises relevant for security is that users are now less involved with the model training process themselves.
Moe HaidarGUEST
23:44
However, for some specific use cases, we do fine-tune our own model.
Moe HaidarGUEST
23:48
And we use SageMaker to fine-tune our model and deploy it and run it.
Moe HaidarGUEST
23:52
And the reason we do this is because foundational models weren't good enough for what we were doing in this specific case.
Moe HaidarGUEST
23:59
And this is always for me, if I was to advise people, check with foundational models.
Moe HaidarGUEST
26:57
We basically implemented them.
Camille DavidovHOST
26:58
Yeah.
Camille DavidovHOST
26:59
So the Spark is built on different layers from model section you mentioned between using bedrock to use the foundational models or to use SageMaker to train your models, your own models or fine-tune your own models to the orchestration, which is by itself a big topic, right? Investing time in your agents and your orchestration, using the data to do the reasoning.
Camille DavidovHOST
27:25
So all this intelligence was built to different layers that you invested a lot of time, I assume, and And this is always evolving, right? So I assume that it's evolving every day, right? The way that you structure it, the way that you use the new models, the way that you evaluate, because those evaluations are being also changed and changed constantly.
Flo MotlikHOST
2:09
Plus a standard tier for guardrails grounding checks, prompt shortening and advanced prompt optimization, domain and location filters for web search, and GROK, 4.7 in GovCloud.
Flo MotlikHOST
2:21
Three updates for SageMaker A, a serverless model customization.
Flo MotlikHOST
2:25
New training session APIS, let you write your own training loop, while SageMaker manages the GPS, plus configurable sequence lengths and support for Muse Glimmer, QEN 3.8, and Ministral 3.
Flo MotlikHOST
2:38
Neuron 2.33 brings a native PyTorch backend for Tranium, now in public beta.
Flo MotlikHOST
2:44
Quick, the Agentech workplace assistant, launches in the Singapore region with data stored and processed locally.
Network ChuckHOST
9:35
This is the machine learning part.
Network ChuckHOST
9:38
Amazon has a thing called SageMaker, which I love their names.
Network ChuckHOST
9:41
The names are hilarious.
Network ChuckHOST
9:45
That sounds like a cool thing.
Network ChuckHOST
9:50
That'd be cool.
Network ChuckHOST
9:51
Everyone able to hear me? Cool.
Network ChuckHOST
9:53
So SageMaker.
Network ChuckHOST
9:55
This is their machine learning platform.
Mahdi SajjadpourGUEST
24:03
Agent Core itself is over 10 services, so things like memory, gateway, et cetera, and it gives you all the capabilities you need to be able to create an interoperable agentic environment.
Mahdi SajjadpourGUEST
24:15
And then Bedrock and SageMaker for you to be able to actually leverage models.
Mahdi SajjadpourGUEST
24:19
For example, if you want to leverage OpenAI's models, those are available through Bedrock itself.
Mahdi SajjadpourGUEST
24:24
And the Data Foundation, which is becoming critically important in being able to create that interoperable interconnected world.
Colin DunlapHOST
2:32
And he has a lot of skills, had a lot of skills.
Colin DunlapHOST
2:35
Data Build Tools, SageMaker, Claude Code, Data Analytics, Other Coding Skills, SQL, Python.
Colin DunlapHOST
2:45
Prior to that, he had experience that went back 16 years.
Colin DunlapHOST
2:49
The guy had worked at big-time Pittsburgh companies.
Arron BailissGUEST
45:30
So that's what Strands Decider is.
Arron BailissGUEST
45:32
It's one of AWS's System 1 models, hosted on SageMaker, run it locally.
Arron BailissGUEST
45:38
We have various model sizes, I think ranging from 2 billion parameters up to 20 something billion.
Arron BailissGUEST
45:44
And as the parameter count scales, as does the model's abilities, but also the compute requirements for it as well.
Jesse OwenHOST
15:15
If you've wanted to customize a model but stopped at the GPU setup, this launch removes some of that work.
Jesse OwenHOST
15:22
AWS made SageMaker AI training sessions generally available on the 22nd of October.
Jesse OwenHOST
15:28
Working AI builders can write their own Python post training loop, including custom reward logic, while AWS manages GPU allocation, distributed execution, fault recovery, and checkpoints.
Jesse OwenHOST
15:41
That gives the developer control over individual sampling and training operations.

1 more episode mentions Amazon SageMaker.

Create an account to see the whole feed, search across every transcript, and follow the entities you care about.

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.