MLOps Accelerator Pilot
Information Technology & ServicesBusiness Impacts
Potential Long Term Savings
Reduced Development Cycle Times
Customer Key Facts
- Location : USA
- Industry : Technology
- Size : 1000
Evaluate current Data and ML landscape and outline an MLOps framework on Google Cloud (Vertex AI)
The client is a privately held technology company headquartered in Lehi, Utah that develops cloud-based software to help businesses modernize customer interactions, such as messaging and customer feedback, and improve their online reputations.
The client wants to evaluate their current Data and ML landscape and outline an MLOps framework to help them define core processes and technical capabilities, and establish mature ML Ops practices.
Challenges
- Multiple programming languages across different stages of the pipeline
- Memory constraints on the payload for online inference
Technologies Used
Cloud Storage
Vertex AI
Cloud Functions
BigQuery
XgBoost
Solution
Quantiphi team assessed the current state of data architecture, data science maturity, and business priorities, and developed customer curated architecture along with deployment and demonstration of ML Ops pipeline for the shortlisted Churn Prediction use case
The engagement was planned in two phases, each including two sprints - the discovery phase followed by the pilot implementation phase
The discovery phase consisted of conducting sessions with the client to understand the use case and the data- schema, columns, and tables involved.
Post the discovery session, the team prepared a detailed document on the current state assessment and the high-level solution architecture that can be used for implementation.
The pilot phase included architecting, modeling, inference, and model serving along with a demonstration of model monitoring and CI/CD frameworks.
Results
- Developed robust orchestration pipelines
- Automated pipeline to handle KPIs and score tuning
- Created ML Ops framework to address customer churn use case
