Leading an AI-powered transformation with far-reaching impact

Before McKinsey, I was a lead data scientist at a unicorn startup, which builds solutions for industrial applications and the federal government.

When I joined McKinsey in 2016, I was one of the folks brought in to build our tech skills and portfolio. At the time, McKinsey was making a concerted effort to hire data scientists; I was part of the first generation hired. We were still trying to figure out how McKinsey could be a player in data science, and my ideas about doing production grade code and building machine learning models seemed a bit out of reach.

Six years later, that’s not the case. Now, I have a dozens of data scientists who are deploying 400+ models in the cloud and driving $100 million of impact. It’s funny, but when I tell some of my friends who are McKinsey alumni, they almost don’t believe it.

Using data to drive changes at the largest competitive power producer in the US

Ayush Talwar
Ayush Talwar

To reduce fuel costs, Vistra, which services roughly 26 million customers in Texas, partnered with McKinsey to deploy an AI-enabled heat rate optimizer (HRO) that would provide the best energy efficiency at any point in time. Vistra had all the IoT in place, and they were collecting and storing data at all the plants and in the central depositories, but their method of storing the data required heavy manpower for analysis.

McKinsey pulled together a team of power industry experts, data scientists and machine learning engineers to build a multi-layered neural network model to underpin Vistra’s existing HRO so they could glean insights from the data and make it actionable. We helped to build talent, capabilities, and output that could be scaled across their 75 plants.

My role on the engagement was to set up the internal team of 20 technologists, 16 data scientists and data engineers, establish a manner for us to efficiently scale our work maximizing impact, minimize risk, and keep best practices in place. Additionally, I helped guide the clients through the process and the formation of relationships with key technical stakeholders. This was critical to building the change management muscles they needed to drive the transformation.

We developed 400+ models, all of which are in production across the 30 locations and 65+ units. Across the plants, we improved the fuel efficiency by 1-3% using AI-enabled methods. Between 80-90% of the models have gone from proof of concept to minimum viable product to scale, with high user engagement.

While Vistra had strong product ownership and data engineering, we helped them scale and start building a data science bench. As a result, they’ve been able to attract some great talent and build up a strong data science talent base. Applying our models across other organizations

We will be able to replicate our work for other fossil-fuel-based power companies, but what gets me excited is the potential to apply this across many other industries. This model of scaling across a fleet of distributed assets and improving operations, efficiency, and maintenance has implications in all types of manufacturing plants.

We’re now expanding the work into renewables because so many power companies are at the edge of transitioning from conventional to renewable energy. Even within the conventional side, for example, coal, better efficiency means better abatement – e.g., 1.5-2 million tons-worth of carbon dioxide. There is a significant environmental impact coming out of this work, which makes me proud.

Data science at QuantumBlack, AI by McKinsey

People have this perception McKinsey is just nontechnical work, but that couldn’t be further from the truth. McKinsey is at the frontier of enterprise AI. We do high-tech projects that have far-reaching impact. We’re focused on finding the right solution for a given problem. We prioritize value creation for clients through a technical framework.

More about me

Ayush Talwar
Ayush Talwar

At McKinsey, I’m part of the data science guild, which focuses on the future of data science and data scientists at the firm. It’s something I am deeply passionate about. I am also a reviewer for several data scientists, so I collect feedback and help them develop and grow within the firm, which is fulfilling.

I love spending time with my family, and we travel as often as we can to different cities across Asia and Europe.

Find a job like Ayush’s

Never miss another post

Receive new stories once a week directly in your inbox