pipeline
PulseAugur coverage of pipeline — every cluster mentioning pipeline across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Hugging Face introduces decision models for structured AI outputs
Hugging Face is introducing "decision models," a new type of model designed to directly output structured decisions and probabilities rather than generating free-form text. These models are particularly useful for tasks…
-
MLOps Series Explores Productionizing Machine Learning Systems
This series of articles delves into the intricacies of MLOps, focusing on the art of moving machine learning systems into production. The content covers various aspects, including MLOps maturity models, advanced techniq…
-
Language models build quote-grounded healthcare software knowledge graph
Researchers have developed a novel pipeline using language-model agents to construct a knowledge graph of healthcare software adoption, ensuring every data point is supported by a verbatim quote from its source. This me…
-
Hugging Face Tools Explained: Fine-Tuning and LoRA Guide
This article provides a practical guide to essential tools for developers working with Hugging Face. It covers key concepts such as pipelines, auto classes, and datasets, along with detailed explanations of fine-tuning …
-
Developer builds AI pipeline to generate exam questions from past papers
A developer is building a system to generate exam questions from a school's past papers, using an LLM to extract and tag questions to learning objectives. The prototype involves splitting PDFs into sub-question items, g…
-
AI assists in pipeline triage, reducing issue resolution time
AI tools can significantly accelerate the process of triaging pipeline failures by analyzing logs and identifying the root cause. Instead of spending an hour deciphering issues, an AI assistant can provide a plain-Engli…
-
AI Training Data Pipeline: Extracting and Preparing Production Logs
This article details a method for creating a training set for fine-tuning AI models by extracting and processing data from production logs. It covers anonymization techniques, formatting the data for instruction tuning,…
-
Knowledge graph metrics may reflect pipeline, not subject: replication study
A replication attempt suggests that structural metrics derived from knowledge graphs may reflect the extraction pipeline's characteristics rather than the subject matter itself. The author tested this hypothesis by buil…
-
Automated ML Retraining Pipeline Design Guide
This article provides a technical guide for constructing an automated machine learning (ML) retraining pipeline. It details how to manage challenges such as delayed labels, data validation, concept drift, and model perf…
-
Scikit-learn Pipeline Feature Engineering Cheat Sheet Released
This article provides a cheat sheet on feature engineering within scikit-learn's Pipeline. It explains how incorporating feature engineering steps into a Pipeline ensures that each step is fitted exclusively on training…
-
AI optimizes natural gas pipeline operations for efficiency and lower carbon emissions
AI-driven optimization is being implemented to enhance the operations of natural gas pipeline networks. This technology aims to improve transmission capacity, ensure supply reliability, and enable more intelligent sched…
-
LLM extraction audit trails: Versioning models and tracking human corrections
The first article discusses the importance of versioning LLM extractions to accurately track down errors. It emphasizes storing the exact model identifier returned by the API, not just an alias, and including decoding p…
-
AI agent publishing pipeline improved with state machine pattern
An AI agent designed for article publishing faced issues with duplicate content and failed posts due to a lack of shared state and recovery mechanisms. The agent would crash mid-pipeline, leading to orphaned drafts and …
-
LLM email approvals need robust architecture to prevent drift · 4 sources tracked
The core issue with LLM-generated emails in automated workflows is not the model itself, but the approval process, which can lead to message drift if not properly managed. To prevent this, a robust architecture is neede…
-
MLOps pipeline built on OpenShift AI for model drift detection
This article details the construction of a drift detection pipeline using OpenShift AI, a platform designed for MLOps. The process involves leveraging Kubernetes and KServe to deploy a model and then automating the dete…
-
Feature Freshness: The Overlooked Problem in MLOps
The article highlights feature freshness as a critical, often overlooked, aspect of MLOps. It argues that many production machine learning models fail not due to poor model design, but because the features they rely on …
-
Databricks Auto Loader Guide for Robust Data Pipelines
This article provides a comprehensive guide to building robust data pipelines using Databricks Auto Loader. It focuses on implementing essential features such as schema evolution, drift handling, and quarantine patterns…
-
Silent data pipeline failures pose greater risk to AI than loud ones
The author argues that data pipelines which fail silently are more dangerous than those that fail loudly. A broken pipeline that alerts users to its failure allows for prompt investigation and correction. In contrast, a…
-
Eugene Yan details robust testing for ML pipelines and MLOps
Eugene Yan's articles delve into the complexities of testing data and machine learning pipelines, highlighting the challenges of brittle tests and the importance of different testing scopes. He differentiates between tr…