My Journey

My name is Timo, I'm 38 years old and I currently live in Zurich, Switzerland - previous stops include Australia (Sydney, Melbourne), the United States (Miami), the United Kingdom (Chester), and Germany, where I'm from originally (around Frankfurt). I have a lovely wife and two young kids (4 and 2) - yes, that's relevant, as you'll see in a minute :) 

I'm a software engineer by trade, and always had a strong interest in ML - my master thesis was about detecting money launderers with ML, I once co-founded a startup about quant trading with ML, and have been working at Kaggle and Google's Core ML department for the last few years, where I now lead a few teams around ML data management. I think it's fair to say that, when it comes to AI, I know what I'm talking about, and I'm always looking for ways to make somebody's life better with it.

So how did I get into creative writing? 

That's where my first kid comes in, well, sort of. When my first paternity leave was approaching, I decided I wanted to use some of it for a personal project (next to cuddling and being vomitted on by my little one, of course). I wanted something completely outside of my comfort zone and had the crazy idea of writing a novel. How hard could it be? Needless to say, I completely underestimated the work to plan this whole thing, write it, edit it again and again, and then edit it some more. I'll leave the details of this for a future post. Let's just say after 3 months that turned into about 1.5 years, I had a manuscript in my hands that I was ready and eager to get out into the world. Except that I completely underestimated that part as well, and no agent or publisher wanted to buy what I had to sell.

What does this have to do with AI? 

Well, that's where kiddo and paternity leave number two comes in. After experiencing how the publishing industry works and how hard it is to get your work published - especially as a first author -, I decided there may be a opportunity to improve things with AI. Traditionally, new authors send their unsolicited manuscripts to a whole lot of different agents and publishers in the hope that somebody likes it enough to take the risk and publish it. Those publishers often receive hundreds of manuscripts a week from this so-called slush pile and can't possibly spend the time to thoroughly read and assess them all. In reality, a junior editor or even intern spends about 5 minutes on the work that authors write over months or years and discard it in 99.9% of all cases. Needless to say that this is far from being a fair, objective and efficient process. 

So I thought: What if we had a ML model that could predict how sales-worthy a manuscript is (or interesting or groundbreaking - whatever a user wants to optimize for)? Such a model could assess a full manuscript in seconds, make decisions free of emotion, reward great content on its quality alone, and ultimately make both authors and publishers happy. I spent my second paternity leave building such a model.

So this is how writing and ML came together for me - as an opportunity for myself and, who knows, maybe an entire industry. I'll share more about the topics I touched on above in various follow-up posts. Stay tuned!