CTO Sessions: Mark Cusack, Yellowbrick Data

Name: Mark Cusack

Company: Yellowbrick Data

Job title: Chief Technology Officer

Date started current role: October 2020

Location: San Diego, CA

Mark Cusack is the CTO at Yellowbrick Data, which offers a high-performance MPP data warehouse for the hybrid cloud. Prior to joining Yellowbrick, Cusack was vice president for data and analytics at Teradata, with lead product management responsibility for the data warehouse and machine learning portfolio. Cusack holds a PhD in computational physics from the University of Newcastle, UK.

What was your first job? Washing dishes in a pub kitchen in Leeds, England. From there I graduated to working at a RadioShack and being a temporary postman. My first full-time job out of university was with the Defence Evaluation Research Agency, which was part of the UK Ministry of Defence, where I conducted research into parallel and distributed simulation.

Did you always want to work in IT? No. Up until my senior year at high school I wanted to be an architect or a physicist. I’d already visited a few architecture schools in the run up to applying for college. This was around the time of the excitement in physics and chemistry over the supposed cold fusion phenomenon, and some friends and I persuaded the head of the high school science department to let us stay in the lab over the weekend to try to reproduce room temperature nuclear fusion. While we were unable to confirm the results of the Fleishmann-Pons experiment using our homebrew neutron detector, it was so exciting to believe we were at the cutting edge of science that I decided that physics was the way to go.

What was your education? Do you hold any certifications? What are they? I did my bachelor’s degree in physics at Newcastle University. I’d always been interested in programming, so I supplemented my physics courses with computer science courses. At the end of my first degree, I decided that I wanted to continue my computer science studies, which led to a master’s degree in that subject. I was fortunate to be able to combine my love of physics and computing and go on to complete a PhD in computational solid state physics in a little under three years at Newcastle University. I used large distributed computing clusters to model the electronic and non-linear optical properties of semiconductor quantum wells and dots based on some novel theoretical approaches. Though, I did get into trouble one time when I thought it would be a good idea to run a Monte Carlo simulation over all 700 workstations spread across the university without asking. One of those “easier to ask forgiveness” moments.

Explain your career path. Did you take any detours? If so, discuss. After my PhD I spent several years working in academia. I then joined the world of applied research at the UK Ministry of Defence, where I worked on parallel simulation, pervasive computing, grid-based computing, image recognition, text analytics and database system development.

Databases were to blame for the next career detour. The MoD team had developed some really novel in-memory compressed database IP which was spun out to form a startup, and which I joined as a founding developer and ultimately became Chief Architect. We built this up into a business, offering a data warehouse archiving product over 10 years, at which point the company was acquired by Teradata.

While at Teradata, I took an offer to join the Teradata Labs leadership team. This ultimately led to becoming the VP overseeing the data warehouse and machine learning software business.

I became CTO at Yellowbrick Data in October 2020. With a foundation in academic and government research, software and architectural leadership at a startup, and then executive positions at a large corporation, being a CTO was the natural next step for me. 

What type of CTO are you? I’m a completely customer-focused CTO. The best days are those where I get to meet with prospects and customers to discuss their challenges and work out how Yellowbrick can help. I learn so much from speaking with customers about how they deploy data and analytics to support their business objectives, and these interactions are critical to ensuring that we “build the right thing” and “build the thing right.”

The next best days are those spent thinking about product vision and how to identify, and hopefully even influence, what the next technology wave will be. I’m also in the rather unique position of leading both the product management function and the systems engineering team. This helps to ensure that the Field is aligned in terms of what’s on the truck to sell and what’s on the roadmap, and the product teams remain aware of how the product they build is being used.

Which emerging technology are you most excited about the prospect of? The growing maturity of container-based software development and deployment most excites me right now. Microservices architectures make it much easier to scale and to achieve resilience and they allow for the rapid addition of new capabilities with zero downtime. Container orchestration frameworks, such as Kubernetes, can run on public clouds and on-premises, providing the ability for the same application to be easily deployed anywhere - even at the network edge. Kubernetes will be a key enabling technology for future distributed clouds, which will provide a unified, federated platform for computing across all deployment types.

Are there any technologies which you think are overhyped? Why? The capabilities of data lake technologies such as Hadoop have certainly been overhyped and have underdelivered in recent years. Unfortunately, this trend seems to continue, but with other technologies replacing Hadoop to support data lake implementations. The problem is that the data lake is often positioned as all things to everyone. On the one hand, it’s a centralised platform for storing and processing large amounts of structured and unstructured data in large file open formats, and on the other, it’s positioned as a platform for high performance data warehousing. These use cases aren’t really compatible, and compromises end up being made. The business ultimately suffers when people try to use a data lake for everything. I’ve witnessed this too often.

What is one unique initiative that you’ve employed over the last 12 months that you’re really proud of? I’m most proud of the distributed data cloud initiative that we launched at Yellowbrick recently. We believe that distributed data clouds will emerge to enable data and analytics services to be provided everywhere in a seamless way: in the public cloud, on-premises, and at the IoT network edge. A distributed data cloud is a federated collection of data management and analytics services that can be delivered at the point of need, and are provisioned, governed, and secured across a single control plane. We have developed a blueprint for how different partner technologies from across the ecosystem can be combined to realise a distributed data cloud.

Are you leading a digital transformation? If so, does it emphasise customer experience and revenue growth or operational efficiency? If both, how do you balance the two? As a rapidly growing business, we’re in a constant state of transformation. We are heavily into automating processes and applying intelligent tooling throughout our business. This isn’t just on the software development side, where we’ve established mature CI/CD pipelines and automated testing processes, but even in our sales, marketing, and customer support efforts too. We have a heavily instrumented and methodical approach to how we generate MQLs, convert them into SQLs, and carry them all the way through to a sale. If something isn’t working, we can course correct rapidly. Our customer support analytics mean that we can detect and fix 40% of problems before our customers even realise they have an issue.

What is the biggest issue that you’re helping customers with at the moment? We’re mostly helping customers navigate their way out of situations where their current data warehouse platform is delivering poor price and performance. Usually, customers choose Yellowbrick after finding themselves in one of three situations. First, they are on legacy data warehouse technology that is too inflexible and expensive to maintain, and they want to migrate to a modern data warehouse platform. Second, they are not getting the performance needed out of their data lake for their SQL workloads and so they bring us in to augment their data lakes. Third, they have adopted a cloud data warehouse but are now facing unpredictable costs and large bills and want to take advantage of the predictable price/performance boost they get with Yellowbrick as an alternative.

How do you align your technology use to meet business goals? Our business goal, not surprisingly, is to grow ARR by acquiring more customers and expanding our footprint with existing customers. We do this by providing a data warehouse with the best price/performance on the market and the ability to deploy anywhere - on-premises and through public clouds. From a technology perspective, we achieve this through our MPP SQL database software, which we built from the ground up using the latest software development approaches, and through a deep understanding of how to get the best performance out of the underlying hardware. Maintaining product quality and delivering business value is extremely important to us. I think we’re doing a pretty good job of it, given our customer NPS of 82.

Do you have any trouble matching product/service strategy with tech strategy? Our tech strategy serves our product and business strategies. We’re always looking at how we can maintain our price/performance advantage. Our decision to adopt a container-based software approach orchestrated by Kubernetes simplifies things, enabling us to deploy our software anywhere - in public clouds, on-premises, or even at the network edge. From a product strategy perspective, this opens up new use cases and expands our addressable market. Our tech strategy sets us up to be the best data warehouse for the emerging distributed cloud.

What makes an effective tech strategy? One that is completely in tune with the product and business goals. So much of the success of a product is due to timing and how effective your go-to-market strategy is. The tech strategy has to be one that ensures that you can bring a differentiated offering to market at the right time, and one that will support a compelling user experience that makes it easy to try, buy and grow with the product.

What predictions do you have for the role of the CTO in the future? A CTO of the future will be one that is not only a technologist, but someone who is responsible for maintaining the alignment between the business, product, and technical strategies. A really close alignment with product management is important here. I’m fortunate to own the PM function at Yellowbrick, so I see the benefits of this close coupling of strategies every day. It’s important for a CTO to critically examine all product and technology investment decisions from an ROI perspective.

What has been your greatest career achievement? Taking a small startup and growing it to the point that the biggest banks and telcos were relying on our product to help run their businesses, and then getting acquired by Teradata. It’s early days for me at Yellowbrick, but without a doubt, we have a very talented and hardworking team here. I’m excited about the direction that Yellowbrick is taking and feel fortunate to have the opportunity to help navigate the company along that journey.

Looking back with 20:20 hindsight, what would you have done differently? The data warehouse archiving product we created at our startup, RainStor, was ahead of its time. As part of the solution, we’d developed a parallel SQL query engine and compression technology that could be deployed in the cloud and on commodity hardware. Our cloud offering stored its data in AWS S3, and RainStor clusters could be spun up on EC2 and query the data in S3 directly. This was a few years before SQL query engines like Presto and Impala came around, as well as data warehouses like Snowflake, which actually has a pretty similar architecture to what we had with RainStor. Timing and GTM is everything.