To address the challenges in machine learning delivery, Capital One’s Senior Manager for Next Generation Business Strategy, Aimee Bechtle and Software Product Innovator, John Schmidt, discussed how they formed a cross-disciplinary team with associates from data science, data engineering, DevOps and product management for a DevOps Dojo at the DevOps Enterprise Summit 2018
In order for Schmidit to embrace APIs, security, opensource and microservices, Capital One needed to provide its DevOps teams with continuous integration, continuous delivery and cloud computing.
This is where Bechtle came in useful, supporting Schmidit with the bank’s investments by bringing a strong hypnosis to the table and guiding him to face the challenges the bank was facing at the time.
‘Turning the company around’
Bechtle decided that working with the internal coaching programme Dojo would benefit the bank, and it appears that she was correct to follow that path.
She revealed to the attendees that implementing Dojo completely helped towards Capital One’s real learning problems; created a psychologically safe environment where teams were able to learn new skills, practices and challenges faster than they would on their own; as well as brought expertise and resources to the team in the duration of six weeks!
Nevertheless, she noted that it was challenging to upskill SW engineers and implement a CI/CD pipeline for each web API that supports federal code contributions and executes the same code.
This is why, together, they decided to work together for months to create two teams of Java Web API professionals, hire 9-10 engineers and focus on DevOps skill sets.
“We sat next to each other every day for months to plan our teams in order for us to turn the company around in a matter of 6 weeks. This had its ups and downs, but we could either quite or move forward,” revealed Bechtle.
‘A complete culture change’
“At first, things weren’t working how they should be, but by the end of it, our tests which previously took 3 weeks only took 3 hours.”
According to Schmidit, the 3-hour turnaround wasn’t just achieved because of the new skill sets, but, instead, over a complete culture change.
After they finally achieved their goals, they decided to move towards machine learning and the container orchestration platform. Interestingly, a similar pattern emerged, with 6-weeks being the minimum time in order to implement new patterns, practices and to make skills stick.
“For this new journey map to be a success we called in an agile coach, studied previous survey results, and created different exercises for our teams,” admitted Schmidit.
Both experiments proved that DevOps and Dojo are two huge speed levers. Reusable building blocks increased; their options continued to create a durable competitive advantage; and people, finally, had more control over their money – in one case, the speed of filtering went from months to hours.
“We needed to understand our audience and what motivates them”, he added.
According to Schmidit, to support a change to DevOps in your organisation, you must follow the below steps:
- Look for the innovative, forward-thinking product manager
- Engage both product management and tech
- Know what motivates them
- Frame your arguments around the outcome
- Form a product leadership team for the journey
- Leverage a Dojo, immersive learning model to accelerate the change
- Have product leadership motivate and support teams to get through dips
- Avoid the narrow perspective, while looking beyond your context
- Celebrate success!!!
He also noted that, as a Product Manager, he is hugely interested in investing in software to help create and sustain a competitive advantage, but, sometimes, options can be too constrained.
Written by Leah Alger