Introducing Suresh Vadakath, Arize’s Senior Solutions Architect
Arize AI is currently hiring over ten positions. Join us!
Suresh Vadakath is a Senior Solutions Architect at Arize. He brings over a decade of experience spanning both consulting and technical client-facing roles, including at Dataiku, DataRobot, and Alteryx. Suresh lives in New York City.
Can you introduce yourself and share your career background?
I started my career working in consulting on supply chain system integrations. Around 2004, I went to graduate school at the University of Rochester’s Simon School of Business, focusing on investment finance. At the time, I wanted to be an equity research analyst but decided to get back into tech in the aftermath of the 2008 financial crisis. In retrospect, this was a good move as it combined my skills from graduate school with prior tech experience. My second inning in tech landed me in consulting and a few years later, I got approached by Alteryx to be their first Sales Engineer in the New York City metro area. That opened up an entire new arena of enterprise software sales. Terms like go-to-market were foreign to me back then. I then worked at a few other companies – including Dow Jones, DataRobot, Dataiku – before joining Arize.
Why did you first decide to get into machine learning?
It felt like the natural next step to transition from analytics to machine learning. It’s a very innovative space with a mystique attached, so that was attractive from a continuous learning perspective. ML is also more and more fused with decision making. All of that coincided with data science platforms and content to level-up becoming more accessible and mainstream, making the transition easier.
How would you describe your role and responsibilities at Arize?
I’m a pre-sales solution architect, so I’m a technical counterpart to an account team to build Arize’s footprint. On the customer-facing side, this involves presenting our platform as well as ideating on and prototyping ML observability examples and integrations into the customer’s environment. We also act as liaisons from the field for user feedback and the machine learning landscape. It’s a multidimensional role.There are also the entrepreneurial, content engineering aspects to it as well which I like.
Why is ML observability so important?
One has to have a risk management perspective about ML observability so you can secure and grow your investments in ML. Lot of factors degrade the quality of ML model predictions which requires ML teams to properly fix issues in a timely fashion. The task at hand gets amplified when you have high value use cases, growing prediction volume, a bigger model portfolio, and large teams. So resolution time, precision, richness of insights, and scalable architecture all matter. And doing that for scenarios where structured and unstructured data elements are involved is what a good ML observability platform ought to provide for in a centralized place.
What’s one thing that has surprised you since joining Arize?
How many cool outdoorsy activities my colleagues partake in during the weekends (I can’t keep up). Jokes aside, the abundance of talent was not a surprise. There’s a laser focus on execution, including competency and customer delight. Importantly, it’s being done in a supportive environment for colleagues, customers, and partners.
You focused on financial services in your prior role and are one of the few data scientists I’ve met with an MBA. What do you think are some of the unique challenges facing financial services companies deploying models into production today?
It depends on a few factors including user case type. Asset price prediction for trading purposes is a use case in an extremely dynamic domain so there are a fair share of skeptics and adopters of ML or present-state based techniques. But just like with anything else, there are adopters (systematic quant) and discretionary side teams using ML as a reference tool or to just understand relationships between things without deploying models. Recently, I was reading about a firm using a technique called meta-labeling to make inferences on size of trades based on the probability scores of a secondary classification model that learns from a primary model where the position (aka label) to go buy or not is set. This stands in contrast to simply using ML to come up with the investment strategy. Take natural language processing NLP) as another example: it’s one thing to classify or extract unstructured data, it’s a whole other thing to do that in an automated investment decision context, given that it’s a low signal and high noise domain.
In general, there is a lot of data appropriate-ness and preparation that come into play as well as regulatory scrutiny on the process for things like fairness and bias in financial services – especially with credit decision flows. Nevertheless in my experience, I’ve seen more ML deployments on the client-facing side in consumer credit, wealth management and back office operations like fraud, trade failure, where the system is in a relatively steady state.
You’ve seen a lot of flavors of infra and data-centric AI from Alteryx to Dataiku and Datarobot. Any interesting lessons from those roles that inform your work at Arize?
ML observability is a very specialized area. Each ML deployment requires a system-level thinking because problems cascade. It requires us to have discussions around environments that are ancillary to Arize (e.g. deployment and integration options in terms of cloud, inference stores, and use cases). We go in-depth with these topics. Aside from the technical stuff, we operate in gray areas constantly so communication is key. There’s a lot of reconciliation of doubt, action and coordination involved.
What is the one app on your phone that you can’t live without (bonus points for naming any underlying ML-powered systems)?
NYT Cooking app is something that my family and I use for quick meals. I’m not sure whether it has an algo, however – there might be some combination of user behavior and collaborative filtering. I remember reading an article about how they test the recipes obsessively.
Since you’re a New Yorker, what’s your favorite pizza joint?