On MLOps Challenges and Pain Points. In previous DM issues, we discussed all the many moving parts of MLOps, a why it’s challenging to synch all the bits & pieces when building and deploying an e2e MLOps platform in production.
How does anyone do MLOps, What are the unaddressed challenges, and What are the implications for MLOps tool builders?
A team @BerkleyUni asked those questions to a group of ML Engineers in interviews across several organisations.
They’ve recently issued Operationalising Machine Learning: An Interview Study In the paper, they publish the results of the interviews, and also review MLOps pain points, anti-patterns, success factors, and common practices.
The guys @INNOQ keep adding great resources and content to the website MLOps -Machine Learning Operations - check it out
The w/e long read. Increasingly, a lot of people in the ML community are researching how to combine ML and Causality. Still, afaik, there aren’t too many real life, practical, success examples out there. In this article Is Causality the Missing Piece of the AI Puzzle? the team @QualcommAI argue that the fundamental open problems in AI are related to the issue of causality, and also show how causality can improve ML.
Have a nice Sunday.
Causal Learning for Accurate Forecasts @Doordash
Rethinking Stochastic Gradient Descent’s Noise
Productising Large Language Models
Does This Button Work? Bad Recommendations @Youtube
How to Architect MLOps on the Lakehouse
Interviewing for a Senior ML Engineer Position
E2E ML Pipeline for Recommendations @Slack
[free] Causal Inference Bootcamp
[free course] Serverless ML & MLOps Principles
Challenges & Lessons: Ads DL Prediction Models @LinkedIn
Share Data Machina with friends
A List of Annotated PyTorch Implementations of NNs
Stable Diffusion Implemented in Tensorflow/ Keras
Automatic Differentiation in 26 lines of Python
Update your MLOps Pipeline with Vetiver & Quarto
Model-agnostic Explainability for Survival Analysis
Missing Data Multiple Imputation with Automated ML
Deep Learning with Flux on Pluto IDE
Lagging a Variable in Time with Dataframes
[Tutorial] Comparing Models Predictions
Twelve Reflections for Better DataViz
Open Healthcare Access Maps per Country
US & UK are Poor Societies with Very Rich People
Google TensorStore for High-Perf, Scalable Array Storage
Build a Reactive, Streaming App with Python & Kafka
100 Billion/messages per Day with Apache Pulsar @Tencent
On the Paradox of Learning to Reason from Data (pdf)
A Generalist Neural Algorithmic Learner
Pretrain, Personalized Prompt & Predict Paradigm for RecSys
[Free book] Bandit Algorithms (pdf, 582 pages)
Scalable Synchronisation Algos & Lock-Freedom
Ant Colony Optimization for the TS Problem
Automating Fast Food Restos with Robots
Using Drones & ML to Detect Land Mines
Nanocopter AI Challenge: Teams’ Presentations
Reinforcement Learning for Beginners
[Tutorial] Graph NNs for Recommender Systems
Transfer Learning Using Pre-Trained Model Repos
Zone7 - AI for Athletes Performance
Hour One - AI for Virtual Human Presenters
Pano - AI for Fast Wildfire Detection
Meter-ML: Methane Emissions Dataset for ML
Global Daily Co2 Emissions Dataset [+ dataviz]
A Dataset of Cryptic Crossword Clues
Tips? Suggestions? Feedback? email Carlos
Curated by @ds_ldn in the middle of the night.