Hi everyone 
I’ve just released a hands-on AI project that brings predictive automation to technical support pipelines:
GitHub: AI SLA Predictor for JIRA
What It Does:
This tool uses transformer-based NLP and historical ticket data to:
- Predict SLA breach risk in real-time as tickets are raised
- Automate priority tagging and routing before escalation happens
- Integrate with JIRA workflows to trigger preventive actions
Built with:
- Hugging Face Transformers + scikit-learn ensemble methods
- JIRA REST API for ticket ingestion and feedback loop
- Flask/FastAPI endpoint for real-time prediction
Use Case:
Designed for support engineers, SREs, or DevOps teams handling high-volume JIRA queues with strict SLA contracts. It works as a preemptive smart layer — especially helpful for detecting risk before a breach occurs.
Feedback I’m Seeking:
- Best transformer models for short-length IT tickets (e.g., BERT vs DeBERTa)
- Ways to improve classification confidence on limited labeled data
- Thoughts on multi-label vs binary classification in SLA breach detection
- Any experience with deploying similar ML models directly into JIRA or Helpdesk pipelines?
This is part of a broader toolkit I’m building to modernize support operations using AI/LLMs — would love your insights!
— Arooj Javed