Kicking off an AI project
25 February, 2019
In Mid of 2017 we, the board, assessed the idea of empowering the platform with AI capabilities. We used a data analytics tool already which allowed us to provide customers linear regression and forecasting. However, we lacked the capabilities of transforming unstructured data into structured ones.
The magic word was NLP, Natural Language Processing. The main difficulty was that we had no AI specialists in our company so far. Nonetheless, I accepted the challenge and drafted an own agenda with two directions. Hiring is something special, as the offers we received from recruiters were either pure data scientists or pure developers CVs. If you are located in Germany, like our company is, the major tech capitals are Berlin, Munich or Hamburg. But you have to cope with a high churn rate, so it's not a bad idea to look also to other major cities with a decent education in there, e.g. Leipzig, Cologne or Nuremberg. All of those cities are connected via high speed trains and have international airports, so ramping up a new office is fairly straightforward - Co-Working spaces exist there as well. While hiring was ongoing, using connections to the local University alumni and selected recruiters, I constantly reserved 4 hours per week to progress reading in scientific papers, books (can recommend the AI book from Stuart Russel and Peter Norvig) and tech blogs to get a good understanding of the domain, tools and frameworks. After 4 Months we had the team ramped up and started straight with a lean prototype which was a chatbot, which has to understand various commands from the procurement domain.
Focusing on a small scenario has the benefit of better testability (no world machine!) and a smooth feature development on-ramp path. We reached out after another 3 months, while having a first showcase ready, to the first initial customers to prove our product vision and get enough feedback for a Beta. It took another 3 months to get to the Beta level, where we had another meeting with the customers. Here we finalised the agenda for a going live project. It is of utmost importance to start now shifting gears from feature growth towards quality and robustness. With this lean approach it was possible to kick-off an entirely new development within less than 1 year. My takeaway, also for future research topics, balance hiring and knowledge gathering can be a very effective way to boost the ramp up time of new technologies within your company, but never forget to hear the customer voice as earliest as possible. It's the only indicator for success at the beginning.
Sameer Kalburgi, VP of Engineering at Fieldwire, discusses how his team evolved from hiring junior engineers to building a team of -- and balancing -- mixed seniorities.
VP of Engineering at Fieldwire
Sameer Kalburgi, VP of Engineering at Fieldwire, shares how he enhanced his public speaking skills to make his presentations more compelling and engaging for a non-technical audience.
VP of Engineering at Fieldwire
Kevin Perko, Head of Applied Research and Data Science at Scribd, explains how standardizing the interview process -- and particularly the interview questions -- resulted in better hires and skill set fits.
Head of Applied Research / Data Science at Scribd
Raghavendra Iyer, Head of Engineering at ReachStack, tells of his efforts to preserve the startup mentality of his team following the acquisition by a large company.
Head of Engineering at ReachStack
Matt Pillar, VP of Engineering at OneSignal, recalls how he helped merge two engineering teams at two different locations and how legal and cultural context made all the difference.
VP Engineering at OneSignal
You're a great engineer.
Become a great engineering leader.
Plato (platohq.com) is the world's biggest mentorship platform for engineering managers & product managers. We've curated a community of mentors who are the tech industry's best engineering & product leaders from companies like Facebook, Lyft, Slack, Airbnb, Gusto, and more.