Choosing a Team: Why I Joined Continual


August 16, 2022

Today marks my first day at Continual and I’m excited to be a part of this transformative journey, one I foresee having far reaching implications within the industry and ecosystem! If I were to highlight my main reasons for joining though, it could be distilled down to three key points: 

  • Strong alignment of core values
  • Having the right focus (empowering data teams to deliver business value without complex engineering)
  • Understanding the overall direction of the industry

Dropping the buzzwords, I would say that it’s an opportunity to work with cool tech that is focused on delivering real world use cases into production - one that works natively with modern cloud data platforms and is aligned with the overall direction of the industry. It’s also being part of a people-centered organization that’s authentic and genuine, one that’s willing to not only say what they do but also do what they say. Finally, it’s joining a company that’s customer-centric in their very nature and genuinely cares about the success of their own customers. 

If you’re ready to go on this journey together, let’s dive into things.

Alignment of core values

We understand that culture and values can be quite personal. For me though, it starts with a few central themes: simplicity, continued growth, sustainability, doing things the right way, and authenticity. Culture can often be described as the ways of working and how we engage each other. Besides sharing certain core values, Continual also has a proven team that is focused on building a lasting and sustainable company - with a mutual belief that there can be a better way. Most of the team can be considered veterans within the industry, with a deep understanding of the technology and market. They’ve experienced first-hand the trials and tribulations of traditional ML platforms. They’ve seen what has worked, where things have gone wrong, and what really matters when it comes to technology, people, or culture.

When I first spoke to the team, you had the sense that you were speaking to genuine people. People that, sure, are smart and unafraid to keep it real with you but who were also nice and cared about you as a person. It inspires confidence that both the founders, Tristan Zajonc (former CEO and cofounder of Sense) and Tyler Kohn (former CTO and cofounder of RichRelevance), have built successful companies before, including exits, in each of their cases. There is of course Jordan Volz, who I’ve worked with previously, but others as well. Very few early stage startups make it all the way and experience matters. Knowing the caliber of team that I’ll be joining at Continual, I have complete confidence in our ability to execute and succeed no matter how winding the path. 

It’s also about the vision. While machine learning and AI has been one of the hottest spaces in recent years, there have been estimates that anywhere between 53% of projects (this is from organizations with some level of AI maturity) to even close to 90% of projects (with those who are still looking to develop a data-driven culture) never make it into production. If an ML use case never makes it into production, is it actually deriving value for your business? 

This brings me to my next point. 

Having the right focus

Everyone is on a journey towards simplification and reduced complexity these days, technology included. There is often too much complexity and too many choices. One thing that I’ve greatly appreciated about Continual, in the clarity of their vision and technology but also mission statement and team, is the focus - a focus around providing scalable, easy-to-operationalize, and practical AI/ML, which is achievable by raising the level of abstraction and having the system automate everything else. We’ve seen similar transformative ways of thinking occur in related fields (such as with analytics engineering), and the goal now is to do the same for ML.

Continual achieves this through their declarative approach to machine learning. This is when I had my own “aha” moment as the “declarative mindset” could be applied to other areas of life, such as what I call “organizational engineering”, i.e. the art and science of building (and scaling) organizations. From my own experience of building out and leading two high-performing global support operations completely from scratch, most recently as the Global VP of Support at Dataiku but also previously at Zoomdata (acquired by Logi Analytics) where I started as support engineer #1, the key takeaway is that the focus should remain on the objective. Everything else can be figured out from there and this will very likely evolve over time. You’ll of course want to consider your particular set of circumstances (which can differ greatly), including constraints, what you are and are not willing to compromise on, what you are looking to optimize, and your specific personnel. However, the end result you’re hoping to achieve drives your approach. 

Answering the age old question of whether the end justifies the means. The answer is yes. 

A personal pet peeve of mine is the notion of moving fast and how it can often distract people from having the right focus, which is one way tech and process debt can start to accumulate over time. “Let’s move fast,” they say (which is true and important) but it’s perhaps more correct to say, “Let’s consistently move fast”. This factors in scale and repeatability when it matters. 

That’s not to say that we should be over-engineering a solution or process either. Process should never be the end goal (Jeff Bezos in his original 1997 shareholder letter famously describes what he called a Day 2 organization, including an over reliance on “process as proxy”). At the same time, small decisions can more often than not have outsized impacts and it would be short sighted to not consider the long-term implications of the decisions you end up making, especially if they become a core part of your business and/or norms in the ways of working.

These traps can ensnare many teams and one shouldn’t lose sight of the end goal or what’s important. In the case for teams looking to implement ML use cases, it’s aligning on a production process, focusing on practicality, and striving for simplicity when possible. In other words, unlocking true and repeatable value from ML/AI requires shifting the focus from typical “data science” work to operationalizing use cases (through software engineering best practices). Bringing things back to Continual, there’s a core understanding of the reduction in complexity, real-world practicality, and ability to put ML into production that a declarative approach can bring to organizations, which underpins the success of Continual and the traction it’s building. 

Understanding the direction of the industry

There’s no doubt that the cloud is going to be a big part of the future and will only grow in prevalence. Some estimates have the cloud market easily surpassing $1 trillion / year in revenue by 2030, meaning we are only scratching the tip of the iceberg. According to McKinsey, one of the keys to success is adopting an agile and cloud-native way of working, with early adopters being able to capture an outsized portion of the total value that is possible and outstripping their competitors. This means not simply “lifting and shifting” existing deployment models, infrastructure, and processes, but completely reevaluating and revisiting your approach. 

Cue the Modern Data Stack (MDS). But exactly what is a “modern data stack”? I like the definition that Barr Moses of Monte Carlo provides to summarize this topic: 

  • It’s cloud-based / cloud-native
  • It’s modular and customizable
  • It’s best-of-breed first (choosing the best tool for a specific job, versus an all-in-one solution)
  • It’s metadata-driven
  • It’s SQL-centric

Honing in on the third bullet point, this implies a level of specialization. However, specialization does not mean complexity. In fact, one could argue the opposite. To make a simple analogy, we can draw a comparison between the MDS and how the human body operates. What do I mean? 

Within the human body, each body part or organ serves an important purpose and they need to work well together to serve a greater goal. Specialization serves its role, and it’s also the seamless, well-thought-out pairing of parts that enables the body to be successful. Sure, things can evolve and change over time but if different parts of the body (or tech stack) are suddenly at odds with and/or even rejecting each other, well we would call that an autoimmune disease.

Taking this analogy one step further, if dbt is the heart of the MDS, then Continual could be considered the brain. When I first spoke to Tyler, he shared a vision where Continual is striving to become the “brain of the enterprise”, or the missing operational AI layer for the MDS. Then, if we extrapolate upon the declarative nature of Continual and how it proposes to operationalize production AI, you don’t “tell” your body what to do. Instead, it instinctively knows the best action to take based on the “end result” that the brain is prescribing. Your body is providing necessary information and inputs to the brain, which then makes a decision, and feeds back a command. 

In more technical terms, Continual is an operational ML platform that’s been co-designed with the modern data stack and powered by a declarative approach (i.e. higher order abstraction), thus enabling those on your data team to more easily build and deploy models into production. This results in a simplification of the process to build, continuously improve, maintain, monitor, and apply governance to your models. Therefore, less time is wasted on infrastructure, gluing together workflows, or trying to work across disparate data teams. 

In business terms, Continual and the MDS being cloud-native helps to future-proof your organization. It enables you and your teams to focus on building use cases and getting them into production, so your business is deriving value from these ML workflows, without having to worry about an eventual migration of your tech stack, your architecture becoming outdated (with increasing difficulty of use, lock-in, and suboptimal solutions being very real concerns), and tech or process debt accumulating. Focusing on the 95/5 rule, your team can spend more time tackling the most challenging (the 5%) of problems facing your business and less on trying to operationalize the 95% of use cases which can be standardized and to a degree automated. 

Getting started and closing thoughts

So how easy is Continual to use? Well, to get the initial setup going, you’ll need to:

  1. Register for your own free trial of Continual here
  2. Create a project and connect to your data warehouse (+ install the CLI
  3. Define the features and inputs that you’d like to use (even easier if you are using dbt
  4. Define a few configuration files so that your models can be trained declaratively 
  5. Use a single command to “push” these config changes to Continual so that your models are automatically built, trained, deployed / promoted, and predictions can be made
    Note: Continual being data-centric means that your predictions will be stored in your data warehouse and made available for further use downstream (or even reverse ETL)
Not as simple as 1, 2, 3, but.. close

However, rather than take my word for it, feel free to request your own demo so you can see for yourself. We also have guided examples, such as customer churn and lead scoring (and more coming!), to help get you started and leveraging declarative AI more quickly. 

I’m super excited to get started at Continual, you can expect more announcements to be coming soon, and I hope to see you join us on this journey. Now, let’s go get it!

Sign up for more articles like this

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Choosing a Team: Why I Joined Continual

Andrew Tsao discusses the main reasons he chose to join the Continual team.

Aug 16, 2022
Modernizing MLOps: Why I Chose Continual

Traditional AI platforms require users to manage complex data infrastructure and write bespoke ML pipelines for each case. We let users build predictive models using a declarative workflow that radically streamlines operational AI at scale. 

Aug 10, 2022
Book a demo