Changing the culture and service offerings of a big consulting firm isn’t easy, but BCG has been on that path for the past five years. BCG has evolved from traditional consulting services into a digital transformation powerhouse with six divisions that deliver strategic and technical services to clients.
One of those divisions, BCG Gamma, is a global team of world-class data scientists who build data analytics, machine learning, and artificial intelligence solutions for the firm’s clients. But building and shipping analytics, ML and AI applications to clients is challenging. Andrea Gallego, CTO of the division, is charged with creating an infrastructure that can support delivering high-quality ML and AI models at scale.
The Challenge: Shipping ML and AI Software to Clients at Scale
The big question on her mind was how BCG Gamma could not only build models, but deliver them to clients at the edge with real-time orchestration, monitoring and updates. GDPR and other regulations also meant she had to do this while ensuring integrity, consistency and lineage across data models.
Andrea and her team launched the BCG GammaX initiative, a core team of 30 engineers specializing in analytics software engineering, data engineering, UX design, distributed systems, and machine learning engineering. Their charter: build the software and systems to support the rest of the Gamma team.
A Fast Build and Ship Process for Complex AI and ML Models
They turned to Docker to help them automate the process of building and delivering AI and ML models to their clients. With Docker Enterprise, BCG has been able to:
- Quickly build and ship customized AI and ML models to clients
- Maintain software consistency between internal development and external customer production environments
- Assure regulatory compliance with GDPR
Docker’s container platform met BCG’s unique requirements for a mirrored software repository and delivery to the edge, and also allowed BCG to customize the implementation to meet their needs.
Today, Docker Enterprise allows BCG Gamma’s data scientists to build and quickly share initial models with clients, run tests, and then ship final models in Docker containers to the client site.
The team created a software supply chain with Docker Trusted Registry to manage delivery with a mirrored instance to control delivery to clients. This also allows the team to verify image consistency even after the software is deployed, which ties back to meeting compliance requirements. With DTR’s promotion model, the team saves time and guarantees the quality of their software by avoiding labor-intensive rebuilds.
With Docker Enterprise, BCG Gamma can also run its models in containers anywhere. They deploy containers on Kubernetes using Docker Enterprise’s integrated support, and currently run internal systems on AWS (but could easily run them anywhere else).
Each application BCG Gamma builds typically runs across 20 containers. With the supporting infrastructure, they have quickly ramped to over 200 containers, and are adding more regularly.
To learn more about BCG Gamma’s journey with Docker, watch Andrea Gallego’s presentation from DockerCon 2018:
Docker has helped BCG Gamma scale to deliver AI and ML applications to their customers. With Docker Enterprise, BCG Gamma engineers can make changes completely independent of the underlying infrastructure and other dependencies. The team can also roll out new tools, or even experiment and test ideas quickly on Docker Desktop, getting feedback in an hour or two. That lets them make changes and develop new applications much more efficiently.This post was originally published on July 26, 2018 on the Docker blog.