Datasets:

atulkrs
/

mlops-devops-sentiment

text string	label int64	domain string	text_length int64
The MLflow experiment tracking made model comparison straightforward.	1	mlops	8
The Docker image size ballooned to 8GB making deployments painfully slow.	0	mlops	11
The monitoring alert thresholds are misconfigured, causing alert fatigue.	0	devops	9
The feature store integration reduced training pipeline time significantly.	1	devops	9
Kubernetes pod crashes are happening every 2 hours with no clear root cause.	0	mlops	13
MLflow tracking server went down and we lost all experiment metadata.	0	mlops	11
GitOps workflow with ArgoCD made deployment auditing much cleaner.	1	devops	9
Docker containerization eliminated all environment inconsistency issues.	1	mlops	7
The RAG pipeline retrieval quality is poor for domain-specific queries.	0	mlops	10
The RAG chatbot on Bedrock reduced support ticket resolution time by 40%.	1	mlops	12
The CI/CD pipeline deployment was seamless and completed in under 5 minutes.	1	mlops	12
The model deployment pipeline failed three times due to dependency conflicts.	0	mlops	11
Our Terraform state file got corrupted and we lost track of infrastructure.	0	mlops	12
Model accuracy dropped 15% after the last data pipeline update went unnoticed.	0	devops	12
The vector database query latency dropped from 200ms to 12ms after indexing.	1	devops	12
CloudWatch anomaly detection alerted us before customers noticed the issue.	1	devops	10
Automated model retraining triggered by data drift kept accuracy stable.	1	devops	10
Our canary deployment detected a regression too late, affecting 20% of users.	0	devops	12
Terraform IaC made the infrastructure rollback effortless.	1	mlops	7
Kubernetes auto-scaling handled the traffic spike perfectly.	1	mlops	7
Vector search is returning irrelevant results for most production queries.	0	devops	10
Our SageMaker training job finished with 94% accuracy on the first run.	1	mlops	12
Our A/B testing framework for models made rollout decisions data-driven.	1	devops	10
Our model monitoring dashboard caught data drift before it impacted production.	1	devops	11

MLOps & DevOps Sentiment Dataset

Dataset description

A domain-specific sentiment dataset containing real-world MLOps and DevOps scenarios labeled as POSITIVE or NEGATIVE. Built to fine-tune sentiment classifiers for technical operations contexts where general-purpose models (trained on movie reviews) underperform.

Why this dataset exists

General sentiment models misclassify technical sentences. For example:

"The pipeline failed silently" → general models often miss the negativity
"Terraform rollback was effortless" → domain context needed for high confidence

Dataset structure

Split	Examples
Train	24
Test	6

Fields

text — sentence describing an MLOps/DevOps scenario
label — 0 (NEGATIVE) or 1 (POSITIVE)
domain — mlops or devops
text_length — word count (added during preprocessing)

Source

Manually curated by @atulkrs based on real-world MLOps and DevOps engineering experience.

Usage

```python from datasets import load_dataset ds = load_dataset("atulkrs/mlops-devops-sentiment") print(ds["train"][0]) ```

Intended use

Fine-tuning sentiment classifiers for MLOps/DevOps tooling feedback
Benchmarking domain adaptation of general NLP models
Curriculum data for MLOps-aware LLM fine-tuning

Downloads last month: 33

Number of rows:

Total file size:

10.3 kB