My accidental business celebrates 3 years with surprising growth
Accidentally building a thriving AI consulting company operating internationally.
It’s fair to say I’ve started my business, ulam.ai, by accident. In September 2016 I finished being a Research Fellow at Oxford and returned to Warsaw to become an assistant professor at Institute of Mathematics of Polish Academy of Sciences. You see, I was a mathematician doing pure mathematics. I was as far from business and practicality as possible back then. In 3 years everything changed.
Business started by accident
Around 2015, while still at Oxford, machine learning picked my interest, while I was looking for a way to automate proving math theorems. On the other hand I was reading tech news regularly so I was aware of amazing success stories of startups (and was naive enough to not know about epic failures of many more). Thus I had a notion of starting and running a business but I found difficult going into that direction as I had a stable position in academia as a researcher in various institutions (Paris, Oxford, Warsaw).
Nevertheless when I got back to Warsaw in 2016 I started talking to friends about potential collaboration. After all many of my old friends went into business and the first consulting gig quickly found me. Without even forming a company I was contracted by a large international enterprise for a couple of weeks. Impressed by my academic CV, they simply wanted to exchange ideas with someone on their next project.
It was a small project and it ended after two weeks, leaving me wanting more. I didn’t know exactly what I wanted to do and how to do it: no presentations, no sales pitches, no business plans. I vaguely knew that I wanted to do some AI/automation consulting and solve business problems to learn more. I was driven by curiosity like in my academic career.
The next gig came faster than I knew. I started a collaboration in March 2017 with a startup formed by a couple of my colleagues. This time I’ve decided to register a company and officially start under the new brand. The company was registered between March and May 2017 and I made an official launch (on social media and for friends) in May 2017, when I had a logo and a basic website. I named it ulam.ai after Stanislaw Ulam, a Polish mathematician working for Manhattan Project and AI being the new atomic energy of our times.
Smart name but that was it. I didn’t know anything about the market I was entering nor how to do anything that is deemed essential to build a profitable business: pricing, sales and marketing plan, hiring et cetera.
Venture building begins
Around the same time I started the first consulting gigs, Enio, a friend of mine, offered me to form a startup around an idea of having anything delivered within an hour. I remember vividly Enio, who was back then a law student and an owner of a speakeasy bar 6 cocktails in Warsaw, explaining to me his ideas over a cocktail. Funny coincidence I’ve just read about a similar venture in India and decided it’ll be a good adventure. Neither of us has run a tech startup before and especially one which would be so complicated when it comes to logistics (driving around the city delivering packages to people). We launched our app in 6 months, end of 2017, using a software house and our own skills. And what a ride it was! But that’s a different story, which I’ve already described elsewhere.
All this — my academic career still going on at the time — meant that ulam.ai had to be side tracked. Bring, my startup with Enio, took off and failed in 2018. Soon after Bohr Technology came to be, a quantum computing startup backed by CDL in Toronto and operating globally from the start. It finished its existence with a splash in 2019. All that took my attention away from ulam.ai.
Hence until mid 2019 ulam.ai existed but I’ve never took any extra time to think about it business-wise: what clients to approach and how, what services I provide and how I price them. It was all ad hoc project based. I was more concerned with building new ventures than developing the one I already had.
Surprisingly from time to time I was getting an email or a call asking for AI consulting, someone inviting me to a collaboration or simply wanting to hire a team of machine learning experts. And whenever I could I took those opportunities, learning along the way and working with both large enterprises, small startups as well as public institutions.
Together with my team (or alone) we’ve been building machine learning models, analysing investments, building AI strategies or just advising on tech related issues. Even though it was a side business, I’ve gathered a rich portfolio of projects in these past 3 years.
AI consulting and strategy in 2020
By mid 2019 I was out of other startup engagements. In September 2019 my contract with Institute of Mathematics in Warsaw finished and I was out of academia, a step which I planned for a long time — always scared to take it.
Starting with October 2019 ulam.ai was my sole venture. Moreover also in October 2019 I moved to London, UK. Suddenly I was in a totally new, yet known, environment and on my own when it comes to business. I’ve started to wonder what I want to do next when it comes to ulam.ai and generally when it comes to AI. My initial dream of automatic theorem provers became a dream to build a MVP for Artificial General Intelligence. My interests stayed through years with text analysis, machine reasoning and reinforcement learning. I was finally free to do something in this direction on my own.
Since June 2019 I’ve been tinkering with some new ideas and it was the perfect time to test them. First came Petacrunch and then Contentyze building on the same set of ideas: text generation and content automation. The initial tests were successful, but it’s too soon to write about them in detail (see you in another 3 years). However they reassured me that I’m going into a good direction. I doubled my efforts related to building a SaaS platform and tackling text-generation problems. Contentyze came to be in January 2020 and the first beta version of the SaaS platform was out in March 2020. So it’s all new and still heavily changing from day to day.
But while while playing around with Contentyze and text platforms, my main course was ulam.ai. By the end of 2019 I knew I wanted to make it into a better company, reworking fundamentals and taking it to the next level.
Additional motivation came from the fact that since Autumn 2019 ulam.ai was my primary source of income. Slowly I managed to secure new recurrent streams of revenue and I had a regular flow of new consulting gigs. Nevertheless a more complex business strategy was still missing in 2019. Ulam.ai was growing thanks to social media and a vast network of contacts I’ve built in the past years.
Time passed between consulting gigs and working on Contentyze.
Here we are, April 2020 in the middle of crisis, and yet ulam.ai has never been healthier. March 2020 was record high when it comes to revenue and April looks similarly well.
ulam.ai became a boutique AI consulting firm by itself, without much of my help or even despite my insufficient skills and knowledge. I ended up with a healthy business which gives me a lot of fun and sustains my lifestyle. Now it’s high time I take ulam.ai to the next level, reflecting on the past and taking bold steps forward.
What’s next for ulam.ai
In the past 3 years, while running ulam.ai without a plan other than ‘do AI consulting’ I came to understand what I like and dislike about it. I love helping others overcome difficult problems be it technical, business or both. I like hard intellectual challenges, especially if they result in tangible results. It attracted me to academia in the first place and I’m constantly searching for that in business as well.
What I don’t like is being a subcontractor, building products for others, providing raw coding and engineering skills. In other words running a software house or a data science shop is not for me.
My research background causes me to value free intellectual tinkering over codified approach to innovation. I’ve always been deeply influenced by Grothendieck’s metaphor of solving a problem. Trying to solve a hard problem is like opening a nut. You can try to crack the nut open using a hammer, hitting it over and over. It’s direct, brute force approach. Or you can try to immerse the nut in a liquid, think about a rising sea, and wait for the nut to crack open. This is indirect approach of building proper frameworks, patiently learning, tinkering with ideas until you find the right one. This second approach is dear to me and it is how I want to run the business. It’s also similar to how Paul Graham views finding startup ideas in his essay.
Having said that there are two ways to take a company to the the next level: productised services and your own product (to elaborate on leveling up companies, read a great essay on wealth by Paul Graham, Ravi Naval tweet series, Nathan Barry’s blogpost; they all discuss the same topic with slightly different angles). You need a way to supply whatever you do at scale.
Building my own product is what I was trying to do all along with Bring and Bohr, but didn’t managed to carry it out. Neither of them complemented ulam.ai activities. It was basically managing two separate businesses at the same time, which doesn’t work well usually (and it didn’t for me, I had to side track ulam.ai). This time, with Contentyze being in synergy with ulam.ai, I am much better equipped, more experienced and with a good plan I believe. And last but not least, with revenue to reinvest.
‘Productised services’ means you have a menu of services to choose from with a fixed pricing. Think about barbers, you don’t negotiate price each time you go to have your hair cut. There’s one price and you know what to expect.
This is exactly what I plan for ulam.ai in the next months and years to come. I want ulam.ai to be an AI consulting company with exactly two offers:
- individual AI consulting
- data science strategy
The first option should be clear, it’s a basic form of consultation, paid per hour for a fixed fee.
Data science strategy offer is what is really unique about ulam.ai — it’s building a unique, in-depth strategy from technical as well as business perspective related to machine learning, data science or AI problems. Whether you’re building a data science team, implementing AI features in your company or applying for an R&D grant, you need a data science strategy and ulam.ai is here to help. I really like building strategies because each case is different, I need to understand it business wise and then think about technical problems as well. And then I have to put everything on paper, maybe manage the first steps of a client’s team, but don’t build the actual solution.
In the past years individual consulting and building AI strategies were the most enjoyable projects and I’ve learned a lot from completing them. This is the direction I want to take ulam.ai in. I believe it’ll provide the right framework for stable, growing revenues, ready to be reinvested in more risky projects and unique knowledge.
Furthermore some materials related to artificial intelligence and business will be productised into books and courses. That’s already happened to some extent. In February 2020 I published Data Science Job, a book for people starting their careers in Data Science. I’m currently writing Artificial Intelligence Business, a book with an overview of AI market, machine learning trends and predictions. I’ve recorded two short courses on YouTube related to data science and I plan to create more content in this direction. Books and courses are a good way to create digital products out of services, by utilizing knowledge acquired through consulting.
Putting it all together, I plan to focus my efforts right now on building a standardised offering for ulam.ai, by creating necessary templates (blueprints for AI innovation and implementation) and integrating tools for smoother work (automation, booking service, outreach, marketing). Having completed that I’ll start hiring more consultants to join me on board while growing the scale of operations. ulam.ai will eventually become a product company in a technology consulting market with a decentralized team of experts.
On the other hand I will reinvest large part of my revenues and time into building Contentyze, a text-generation platform for marketing purposes. That’s why I don’t plan to engage in any new projects which would require from me or my team coding. Contentyze is enough and it complements activities at ulam.ai. Those two projects will take roughly 95% of my work time, as I leave 5% for a general curiosity-driven tinkering.
In the end nothing great happens overnight. The last years were years of slow grind. I expect more years of grinding and building what I like. There’s always a price to pay and it’s up to you what sacrifices you’re willing to make. The most pleasant part of business for me is building something myself, following my ideas wherever they take me and overcoming my fears day after day.
I’m happy with the shape ulam.ai has taken over these past 3 years and there’s really a lot to celebrate as we enter the 4th year of operations.
Cheers to ulam.ai!
Cheers to new projects and intellectual challenges!