Centralized AI deployment environment reduced time to scale ML models by 85%

Built a unique AI deployment environment that allowed multiple business teams to run ML models and scale them across multiple departments and geographies

Business challenges

The customer is a leading CPG company with close to 300 Machine Learning models spread across business functions, however, these models were dispersed and failed to scale across geographies. If a model was developed in a particular market and tested on live campaigns, other markets were unaware of it, making the system inefficient. The customer wanted to create a centralized portal where all the machine learning models from across the geographies and departments could be hosted together ensuring easy visibility, accessibility, and industrialization of the models across markets.

Sigmoid solution

Sigmoid built a unique artificial intelligence deployment environment for deploying machine learning models at scale. The centralized platform allowed various teams to configure and run ML models while enabling them to view, download and share outputs. The user interface (UI) lets users define configurations based on countries, brands, models, datasets, and other parameters. Once the model runs were completed, the outputs were stored in Google Cloud buckets that were made available to data scientists for download. It also generated PowerBI report links for visualization.

Download the Complete
Case Study here

Business outcome

The centralized AI deployment environment allowed the customer to run multiple ML models in production without the need for writing code from scratch. It ensured the faster deployment of ML models across various departments and scaled across geographies while significantly reducing the cost of running models.

80-85%

reduction in time to scale ML models

70-80%

reduction in time to move models to production

2x

reduction in the cost of running ML models

Relevant Blogs

Microservices Architecture Blog Info thumbnail

Microservices-based Architecture: Key to Scaling Enterprise ML Models

ML model training blog thumbnail

Top 5 Model Training and Validation Challenges That Can be Addressed with MLOps

How to Detect and Overcome Model Drift in MLOps Banner image

How to Detect and Overcome Model Drift in MLOps