Todd is the co-founder and CEO of OmniSci. Todd originally conceived of the idea of a GPU-accelerated analytics platform while conducting graduate research at Harvard on the role of social media in the Arab Spring, after tiring of waiting hours or sometimes days for traditional CPU-based platforms to run analytic workflows over hundreds of millions of tweets. He later joined MIT’s CSAIL as a research fellow, under the supervision of Sam Madden and Turing Award winner Michael Stonebraker, before founding OmniSci.
How did the concept for OmniSci come about?
After undergrad, many of my friends took jobs in investment banking. They made a lot of money but were working 18+ hours a day, and that lifestyle just didn’t appeal to me. I wanted to see the world. I was feeling open to adventure when I stumbled across an ad on Craigslist looking for English teachers in Syria. About eight months later, much to my parent’s dismay, I hopped on a flight and landed in the Middle East. While I was there, I immersed myself in the culture and began studying Arabic, which eventually led to a fellowship to study Arabic in Cairo, and ultimately, being accepted into Harvard for Middle Eastern studies. While at Harvard, I began to get back into computer science, a long-time passion of mine and a near-major for me in undergrad. I also enrolled in a GPU graphics course because I couldn’t get into anything else, and then a machine learning course. I thought, “This is so fun, how can I work my thesis into this so I can spend my time programming and analyzing data instead of writing a 50-page ethnography?”
The answer to that arrived in the form of the Arab Spring. I saw what was happening and realized that Twitter was being used as a means of communication, which led me to start harvesting hundreds of millions of geocoded tweets (i.e. tweets where the user shared their location) from the Middle East, which quickly turned into a big data problem of sorts. I wanted to correlate what users were saying on Twitter (for example, whether they were pro- or anti-revolution, and whether they espoused secular or Islamist modes of discourse, as measured by NLP measures of overlap with different representative corpuses I harvested), with their demographic profile, gauged by the neighborhood likely to be home for them based on where they would most tweet, and their social graph, i.e. did they follow government or opposition leaders. I even found interesting behavioral correlates, for example, users who scored highly on discursive and social graph measures of Islamism would tend to stop tweeting at prayer times, which required calculating the exact prayer times for various areas of Egypt, which depends on both the time of year as well as one’s location. I was writing my own code and also leveraging the open source database Postgres, but none of it was fast enough, especially when I added in the geospatial aspect of the tweets (even today, most database and GIS platforms are extremely slow at joining large geospatial datasets). I found myself waiting overnight for query results when I wanted to see the data immediately. I wanted to see where on the map people were, to visualize the status of people, but it would take hours of programming to run the analysis, which was incredibly frustrating. If only I could have had the capabilities that OmniSci has now!
I was lucky enough to snag a cross-enrollment slot in the database systems course at MIT, which was being taught by Sam Madden (now an OmniSci board member) and Turing Award winner Mike Stonebraker. Mike had helped kick off the relational database revolution in the 70s and 80s by building Ingres and Postgres (the same db I used in my research!), and he and Sam Madden had co-founded Vertica, which brought to market one of the first databases purpose-built for large-scale analytics. The design and architecture of Vertica inspired me to think about how GPUs, i.e. video cards, with thousands of parallel processing cores, could be leveraged to accelerate three of the key components of big data analytics: querying, visualizing, and machine learning/data science.
Initial curiosity soon developed into a full-blown passion, and I flung myself into writing the first draft of a database system that could not only harness the parallel power of GPUs to speed up analytic SQL queries, but could also allow for interactive visualization of granular data, aimed at the large geospatial analytic use cases I had stumbled over when writing my thesis. My obsession with building the system was severely detrimental to the rest of course work, not to mention my social life, but at the end I emerged with an early prototype which I named MapD, for “Massively Parallel Database.” Although primitive in many ways, the system could run SQL queries over billions of records in milliseconds with no pre-aggregation, indexing, or downsampling, and also visualize the results in real-time. My professors thought the project was cool, and Sam Madden invited me to join his database group at MIT’s CSAIL (Computer Science and Artificial Intelligence Lab) to continue work on it.
While at MIT, the university’s Industrial Liaison Program would often ask me to demo the system to partners in industry, particularly since the first demo (an interactive Tweetmap showcasing a few hundred million real-time geolocated tweets) was easy to engage audiences around. But beyond just thinking it was a cool demo, I was struck by how many of the representatives from companies like Proctor and Gamble, Verizon, and Qualcomm would immediately jump into questions of how they could use or buy the software. It was obvious they didn’t realize how many feature gaps there were in that prototype, or that it was quite nearly held together with duct tape and bailing wire. However, it was clear that there was a market pain around tapping into the potential insight of the big data sources that organizations were struggling just to store, much less analyze. I gradually built up the courage to form a company around the technology, and left MIT at the beginning of 2014 to focus full-time on MapD as a commercial enterprise. It was tough going early on, and I thought I might have to beg Sam to return to MIT, as we quickly burned through the small amount of seed money we had saved or borrowed from our parents. Thankfully though, we somehow managed to win Nvidia’s Early Stage Challenge in March 2014, complete with a $100K no strings attached prize. That in turn led to an initial seed investment from Nvidia, GV (then Google Ventures), and Vanedge Capital, and we were off to the races.
How was the first year in business?
There’s no other way to put it…the first year of business was crazy. We took the full plunge into starting the company and worked tirelessly from my apartment in Cambridge. I had no idea what I was doing, and my co-founder and I were doing it all—writing code, building demos, taking early sales calls—everything. Of course, after winning the Nvidia contest and obtaining our seed round, we had money to hire a few people, but that in many ways added to the complexity, as one doesn’t emerge from the womb knowing the right people to hire, especially for an untested bet on building a new market.
The first year was also crazy because soon after taking funding and starting to hire out the team, it became clear that my co-founder and I had different visions for the company, which eventually led to us parting ways. He had played a sizable role in encouraging me to make a business out of the technology in the first place, and had done a lot of the early blocking and tackling on the business side, so it was an incredibly tough situation to navigate. To complicate matters further, we had just moved from Cambridge to San Francisco, so I didn’t have easy access to the network I had developed while at Harvard and MIT.
Our investors remained supportive of the company during the transition, although I’m sure they, like me, appreciated the daunting odds facing startups that experience co-founder breakups. Personally and professionally, it was a very difficult year filled with ups and downs, and I wasn’t sure if MapD was going to make it. It was one of those things where we just had to put one foot in front of the other, and eventually we caught a few breaks in landing some marquee beta customers, and we were able to raise our Series A round of funding.
Although I would never want to relive that period, in hindsight, it is clear that it was this “trial-by-fire” that ultimately made me a better leader.
What was your marketing strategy?
In the beginning, our marketing strategy was twofold. We were very product-focused, as we should’ve been, and we combined that with some guerrilla marketing tactics to cultivate buzz (although I don’t think we would have ever consciously thought of them as guerilla tactics; it was just a matter of getting the word out about the cool stuff we were doing). At this point, GPUs were exotic, a novelty, and we did our best to capitalize on that. In the first few years, we were fortunate enough to make the front page of Hacker News at least ten times. Whether it was around benchmarks, billion-plus record interactive demos, or blogs on the deep internals of our system, such as our LLVM compilation engine that could take SQL queries and generate hyper-efficient code for both GPU and CPU, we were generating some serious buzz. When we raised our Series A round, we also got quite a bit of press coverage, including the Wall Street Journal, Fortune, Barron’s, Business Insider, and VentureBeat. We were riding the GPU train—building killer demos, contributing thought leadership articles on GPUs, writing engineering blog posts—doing whatever we could to show what MapD could do. We were lean, focused, and grassroots in our marketing.
How fast did the company grow during the first few years?
Growth was slow in the beginning, as MapD, like most database technologies, took a long time and a lot of engineering-years to get to enterprise maturity, and it wasn’t a SaaS app that was quick to market. Many of our early prospects were inbound leads who encountered some of the marketing tactics outlined above, or saw a demo at a trade show, but it was a struggle to get them to judge us by our disruptive capabilities rather than by what we lacked when stacked up against a typical SQL database or BI platform. As GPUs became more mainstream, we were slowly able to build the right engineering team that allowed us to move faster and fill in the gaps in the product. Little by little, the company grew and we started to land marquee customers, which made the next ones all that much easier to obtain.
How do you define success?
To successfully take an idea and turn it into reality, you have to have a transformative vision and the vehicle (and most importantly, the team) to bring that vision to life. For me, OmniSci was that vehicle. Success isn’t always about making a lot of money—sure, that’s nice, but building an amazing company with great people, a team mentality, and a strong culture is a true measure of success. And for me, as someone who is still very much a product-minded CEO, the ultimate measure of success is when the product you bring to market delivers transformative value for your customers and users. When one of our users at Verizon shared that analyzing their data with OmniSci was like a “muddy windshield was suddenly cleared, and I could see the whole road and all around me,” it was more than just a powerful testimonial to me, it was a form of existential validation that all of those years of stress and toil were utterly worthwhile.
What is the key to success?
From a start-up perspective, you need to build a product that solves a real need (although of course sometimes you have to make folks aware of that need) and find a path to bring that to market. What is a burning pain point that you personally have? Chances are, that pain is shared by others and if the pain is painful enough, people just might be willing to pay money to make it go away. I see too many early founders treat finding the customer problem/pain point to go after as a dispassionate process of TAM analysis. If you haven’t felt the pain yourself, or been close to those who have, how will you know if what you are focusing on is a real pain point? How can you know what it takes to solve it?
One of the other things that is often hard for founders (including myself) to grasp, is that while it’s critical to have an overarching vision, it is also critical to be very looped in, and empathetic, with early customers. If you can do that, you’ll start to understand what they truly need, even if they cannot articulate it, and simultaneously understand the potential barriers to adoption of your technology in a concrete way.
Remember to always think like your customers. Keep solving for real, burning pain points and you’ll be on the path to success.
What is the greatest lesson you’ve ever learned?
Not all that glitters is gold. Being an outsider in Silicon Valley, I learned this lesson early on.
Just because someone has an impressive resume doesn’t mean that they are the right fit for your company. Likewise, some of our best hires have been candidates who did not seem to have all the right credentials, at least on paper. Heavily index on bringing on the smartest and most passionate people you can find, even if they are less experienced; given the right support, they’ll be able to figure the rest out. In short, it’s important to trust your gut when hiring. (I highly recommend Ben Horowitz’s retelling of his hire Mark Cranney at OpsWare in the face of significant opposition in The Hard Thing About Hard Things as an example of this.)
That said, no company can 100% make the right hiring decisions. Hence, it is critical to be able to recognize early on the signs that changes need to be made, even if that just means moving someone to a role that better suits their unique capabilities. These decisions are never easy, but ultimately they make or break a company.
What are some quotes that you live by?
I’m inclined to say that I’m not a big quote person. However, as I thought about this question, one came to mind that I think is particularly relevant to any startup that wants to disrupt an existing market or create a new one (not to mention, it’s greater existential relevance):
“In the beginner’s mind there are many possibilities, but in the expert’s there are few” – Shunryo Suzuki
Disruptive innovation characterizes the rise of most successful startups (Apple, Amazon, Netflix, Intel, Nvidia, Tesla, the list goes on and on), which sometimes is ascribable to classic “Innovator’s Dilemma,” or the rational hesitance of incumbents to abandon profitable strategies, or just the deep psychological tendency for success to shut down creative thinking. I think the latter has a greater role than commonly cited in literature, as in addition to having everything to lose, incumbents also naturally develop some level of “groupthink,” and sometimes downright arrogance around what’s made them successful in the past.
The beautiful thing about a startup is not only that there is often very little to lose or risk by choosing an iconoclastic strategy, but that often entrepreneurs are unbridled by the presumptions and consensus that become deep-seated in the minds of those with extensive experience in a particular field. That of course may lead them to make “rookie mistakes,” which are sometimes fatal, but it also often leads to the breakthroughs that emerge out of tackling old problems in novel ways.
This phenomenon is not just constrained to startups, but is arguably foundational to the process of knowledge discovery itself. For example, when starting my research on the GPU-acceleration of analytics while at MIT, I had countless database experts tell me that either it would never work, or at best, would only be applicable to niche problems. Thankfully, some combination of foolishness and stubbornness kept driving me forward.
As Thomas Kuhn notes in The Structure of Scientific Revolutions:
“Almost always the men who achieve these fundamental inventions of a new paradigm have been either very young or very new to the field whose paradigm they change. And perhaps that point need not have been made explicit, for obviously these are the men who, being little committed by prior practice to the traditional rules of normal science, are particularly likely to see that those rules no longer define a playable game and to conceive another set that can replace them.”
Of course, successful startups quickly transition into incumbents, and they in turn are often disrupted by newcomers. We by no means are at that level yet, but I still have found it critical to maintain a “beginner’s mind” no matter what stage you are at. This leads into one of my other favorite quotes from Jeff Bezos about maintaining a Day One culture at Amazon, which we shamelessly stole as a core value at Omnisci:
“Day 2 is stasis. Followed by irrelevance. Followed by excruciating, painful decline. Followed by death. And that is why it is always Day 1.”
What are some of your favorite books?
I was an anthropology major in undergrad, and for me, Sapiens: A Brief History of Humankind by Yuval Noah Hariri was an eye-opening book. Even though I had read countless theoretical works pointing to the socially-constructed nature of reality, the book drove home in a visceral way the extent to which language, culture, and belief systems fundamentally frame our reality. As a startup, particularly if you are looking to create a new market, you are fundamentally in the business of myth-making. When I say myth, I’m not using it in the sense of something being untrue, but à la Joseph Campbell in the sense of creating a shared (and typically new) world view that both internal (i.e. employees, investors) and external (customers, partners, analysts, press) can hang their mental hats on and re-define market needs, and what critical capabilities products should have to solve those needs.
Dovetailing with the above, I also found Thomas Kuhn’s The Structure of Scientific Revolutions (alluded to in the quote above) illuminating, particularly in how it points out that scientific knowledge itself is deeply socially constructed. If scientific theories themselves are to some extent malleable social belief systems, then how much easier should it be to redefine what a market is, or how a customer should go about solving their needs?
Other books I have found helpful as a founder are Crossing the Chasm by Geoffrey Moore, Play Bigger by Al Ramadan et al., and Zero to One by Peter Thiel. And on a personal front, you still can’t beat the Lord of the Rings trilogy!
Tell me about one of the toughest days you’ve had as an entrepreneur.
Looking back, the day I parted ways with the company’s co-founder (and the weeks that followed) were really tough, and I was unsure of what the future would hold. I frankly chalk up making it through that time to my own stubbornness (which in other scenarios can be a liability), combined with steadfast support and counseling from my ultimate coach and the toughest person I know, my wonderful wife Kezia. Her support, especially in the face of all the crazy hours and stress that a startup entails, has made all the difference.
When faced with adversity, what pushes you to keep moving forward?
I think that all good entrepreneurs are persistent to a fault. In general, we keep charging into the wall until we finally break through (and if we don’t, we get up and try a different way). Some of it is intrinsic—the mentality that despite the day-to-day setbacks, you keep going and never give up. Unless you absolutely go under and there are no other actions that you can take, you keep pushing forward. You almost have to approach building a startup according to the Bushido Samurai code, to live according to the belief that until proven otherwise, your startup is “default dead,” to borrow a phrase from Paul Graham. This sort of mentality should be liberating, not fatalistic, as it promotes a mindset of taking the extraordinary bets and doing the unreasonable and unscalable things that market incumbents find impossible or unimaginable, which ultimately is how the Davids can conquer the Goliaths of the world.
What advice would you give to young entrepreneurs?
Quit while you still can! Just kidding.
But seriously, if the venture you are embarking on is not the product of unbridled passion, and I dare say obsession, then you are destined to fail (versus having a slim but real shot). So while startup life can be fun, and is an opportunity to meet brilliant and passionate kindred spirits, it can also be a particularly grueling marathon, of the type where you are forced to sprint every other mile.
But ultimately there are few experiences in life that can be as rewarding as taking the inkling of an idea, or some itch, and building it into something that makes a real impact. That doesn’t have to be via a startup of course, but startups are a particularly effective vehicle for effecting massive transformation in the shortest span of time. So if you have a burning vision, then drop everything you are doing, drop out of school, whatever it takes, and give it a full-hearted shot. After all, no one on their deathbed wishes they had a few more years to work for the man.