AI Observability for Enterprises - Helping ML & DS teams to continuously monitor, explain, and improve machine learning models (including Large Language Models)
Austin, TX
Joined November 2020
- Tweets377
- Following43
- Followers85
- Likes67
Censius retweeted
News Flash📸
Cerebrium now enables you to add ML observability and explainability with @censius
Checkout out how:
censius.ai/blogs/censius-cer…
#MLOps #MachineLearning
Censius retweeted
Some platforms (as @fiddlerlabs, Alibi Detect from @seldon_io, @VertaAI, or @censius) now provide out-of-the-box monitoring solutions for outlier detection using ML & other unsupervised methods.
Censius retweeted
“User-centered, community-centered, and society-centered,” are the three levels defining human-centeredness when designing, developing and deploying AI technologies. (Credit to @landay )
Censius retweeted
Thanks for featuring my work with @QVeraLiao and all other colleagues at IBM Research. It has been a increasing effort around generative AI, and our work outlines what explainability would benefit users who will be using the models!
Generative AI is taking the industry by storm & seeing how it has become a niche of its own,
How can we make Generative AI Models Explainable?🤔
This paper by @sunjiao123sun_ attempts to make Code-based GenAI Models explainable, let's break it down. 🧵
dl.acm.org/doi/abs/10.1145/3…
Censius retweeted
3 Vs of MLOps: Velocity, Validation, Versioning
- Velocity
Rapid experimentation. New idea to a trained model in a day
- Validation
Systematic model evaluation on scenarios of interest
- Versioning
Version models + datasets, maintain fallback models
arxiv.org/abs/2209.09125
1/
Censius retweeted
Our understanding of MLOps is limited to a fragmented landscape of thought pieces, startup landing pages, & press releases. So we did interview study of ML engineers to understand common practices & challenges across organizations & applications: arxiv.org/abs/2209.09125
Censius retweeted
In a recent blog, @censius highlights 5 best practices for #MLOps with MLflow and walks you through how to get started with #MLflow. 🙌
Check it out here ➡️ lnkd.in/eyfu56cw
#MLflow #machinelearning #censius #linuxfoundation #opensource
Censius retweeted
In practice, a trained machine learning model is never final -- concept drift will inevitably cause a performance decline of a production model over time. [1/10]
Censius retweeted
While ML has grown exponentially, several organizations are still grappling to match the pace at which the ML stack is evolving.
Here’s a mega-thread that covers building an AI stack that involves - data tools, modeling tools, and deployment tools🧵
Censius retweeted
Having the right deployment tools for your ML lifecycle ensures your models not just survive but thrive in the real world.
Here’s how you can build a deployment stack for your ML models🧵
Censius retweeted
A detailed explanation of what caused the issue is essential to solving the problem that triggered a monitor.
AI observability tools like @censius provide this functionality with model observability since it offers a complete dissection of every model decision.
Censius retweeted
Having established a robust AI stack for your data requirements, the next thing would be to have the proper modeling tools.
Here’s how you can build your modeling tool stack🧵
Censius retweeted
Monitoring tools such as Comet, @censius, and @EvidentlyAI automate the process and trigger when pre-configured conditions are met.
@censius also enables creating monitors without writing code, supports customized metrics, and helps in root cause analysis.
Censius retweeted
Getting the right data management tools builds a solid foundation for your model development.
But, how do you go about with it?
Censius retweeted
The complete guide to the modern AI stack by Ayush Patel buff.ly/3zKkTrI
Censius retweeted
Random forests are such a great tool! But do you know how they *really* work?
In this interactive notebook, we'll build a random forest from scratch using a classic Kaggle competition. That way, you'll really understand them! 🧵
kaggle.com/code/jhoward/how-…
