Skip to main content

CSSU-OPT-KAI

info

This project corresponds to AI Platform's Generative AI & Knowledge AI MLOps journey.

Project Goals

The aim of this project is to enable operating assets to improve risk management and deliver more reliable plans to the offshore organization by utilizing Natural Language Processing (NLP) models. This project envisions to automaticaly discover lessons learned from previous safety incidents relevant to activities in the operational plans.

This tool mines the content of Equinor's Synergi database to help determine:

  1. What activities were being performed and where?
  2. On which systems / equipment?
  3. Under which conditions when the incident occurred?

You can find more information about this tool at:

https://statoilsrm.sharepoint.com/:p:/r/sites/dcoedatascience/_layouts/15/Doc.aspx?sourcedoc=%7B86B0ABE1-AF3B-4441-B4C6-BA90F7339D33%7D&file=DCoEKAI_jensam.pptx&action=edit&mobileredirect=true&DefaultItemOpen=1

https://statoilsrm.sharepoint.com/:p:/r/sites/dcoedatascience/_layouts/15/Doc.aspx?sourcedoc=%7B86B0ABE1-AF3B-4441-B4C6-BA90F7339D33%7D&file=DCoEKAI_jensam.pptx&action=edit&mobileredirect=true&DefaultItemOpen=1

https://statoilsrm.sharepoint.com/:p:/r/sites/dcoedatascience/_layouts/15/Doc.aspx?sourcedoc=%7B6F32BE10-B403-448D-B7F7-7C9B4D24F289%7D&file=nlp_meeting.pptx&action=edit&mobileredirect=true&DefaultItemOpen=1

Summary of Results

Data scientists using nlp to find the most relevant reports for operational planning purposes. They run a continuous batch job requiring high CPUs and GPUs.

Project Team

Jennifer Sampson (Data Scientist)
Bjarte Johansen (Data Scientist)
Peter Koczca (Data Scientist)
Terje Elde (K8 Developer)

MLOps Challenges

Large Compute Need

The team needed an ability to deploy continously running ML model workload to process reports.

Run Batch Jobs Contineously

During deployment, batch jobs needed to be run continuously while will process plant reports using NLP models.

MLOps Solutions

K8 Deployments for Processing

The NLP models to process the plant incident reports were deployed using K8 Deployment on AI Platform's Kubernetes cluster.

GitHub Repos

https://github.com/equinor/cognitive_ssu/
https://github.com/equinor/cognitive_ssu-k8s