Aito was the sponsor at the Junction 2020 Connected that happened over the weekend from Friday the 6th of November until Sunday the 8th. This is the first big event of the kind we participated in on the organising side. Let’s just say that it’s been quite an intense weekend, but I decided to write some first thoughts about it right after it ended.

The hacker attitude

Aito hosted a competition during the event. We didn’t have the earlier experience, and it was the first time in the Junction history to be a fully online event. So as you can imagine, we had zero idea how many, what kind of or even whether any of the teams would get interested. The challenge was easy to grasp on the one hand but is a pretty challenging multi-value optimising problem if you really want to tackle it properly. Expectations and fears ranged from “nobody will participate” to “there’s too few of us on our side”, but we decided to go with a set three-person task force, consisting of Maria, Tri and myself. We received some interest in the form of meeting requests also during the week, which was a positive sign, but still, we didn’t really know what to expect.

Aito team was waking up and ready for day number two.

From the very start, it became clear that at least our fears of being forgotten in some corner were wrong. In the first few hours, we met with teams from Japan, South Korea, Italy and Hungary, to name a few. The teams were diverse when compared to one another, but also internally, as some of them spanned multiple countries and people who’ve only met hours before. That really was an amazing thing to see, how fluently all of this just seemed to happen. At this point, we already became aware that it’s not going to be a leisurely weekend hanging around by the computer. What more, the teams were really enthusiastic, optimistic and really quite “out there” when it came to the ideas. That enthusiasm and atmosphere caught up with us as well, so it really felt this was going to be something to remember.

Learnings on the go

An excellent choice we made early on was to host our own discussions on our own slack. It turned out to have been a lucky choice since there were some problems with the organisers’ apps during the event. On Saturday, for example, Cloudflare was down for maybe close to an hour. This caused us to miss a few of our meetings, but largely we were able to cope in our own little Aito-bubble. The added benefit was that on the organiser’s platform, you only got notifications for some of the messages, and you could not spawn off discussions as easily to a new channel as on slack. Overall, the systems worked quite well, so the problems weren’t really severe, and to some extent, I guess an event with 2000 people all over the world is quite likely to encounter a glitch or two.

Another learning experience was hosting all of this on our own toolset and platform. Competitions built on OSS tool sets have a clear advantage of being out there, so there’s already a wide set of instructions, FAQs and gotcha’s out there. Mentoring is more about the problem statement than about the tool usage. This became clear really soon, as we received a lot of questions about both Aito usage in general, as well as about detailed specifics about the query API. It’s not often we have some 25 teams do a head-first dive into our system over the course of a single weekend. I’m pleased about our team in this regard, though, since Tri and Maria were handling this with grace. The panicking was left mostly to me. Overall I’m happy with the job done here. Let’s see what kind of feedback we get from the teams. The advanced level of implementation some of the teams reached was really a good testament both to their ability to innovate and the fact that the knowledge-base we’ve created so far is sufficient to get quite far. Not perfect, though, so we have clear ideas about what concepts and details need more attention from our side.

The secret fourth member of the aito.ai Junction team.

So many winners!

Finally, on Sunday, it was time for the judging and scoring part of our journey. We had reserved some ample time to rest in between the days for ourselves. Well, or at least we didn’t pull any all-nighters here. The suspense did make it quite hard to catch sleep in the evening for me at least. The teams, on the other hand, did apparently not consider sleep a part of their Junction-journey, since we received messages from one team or the other throughout the night(s).

I had been a bit sceptical about some of the plans the teams presented, but boy was I wrong! I have to say I was really awe-struck by some of the demos on Sunday morning. Yes, there were missing tests, glitches here and there plus some features were just ideas for further development, but overall the level that some teams reached was really amazing.

Top three aito.ai challenge hacks of Junction 2020 Connected

During the event, our top 1 team had implemented a Telegram bot, in which they’ve thought about things like fairness within the ML-prediction, a functioning secret ballot system and a reconciliation algorithm for the results. And all this in a live demo we got to try ourselves, after some super late bug fixing. Great job, Team FoodVote.

The second price went to Team Brunz. They were the most advanced in using Aito in their solution and had this been a week-long hackathon they would probably have worked some more magic there. We also had some really good discussions during the weekend, so overall this was both a really pleasant co-operation, and the result was quite spectacular. In the end, we didn’t get to try it out, so the last few finishing touches separated them from the first place.

The surprise third prize was awarded to the Team Lunch Match. They’ve done some next-level stuff with their native iOS-app over the weekend. A quote from our review session was, that “do we need an iOS-app like this for Aito? If so, these could be our team for that”. The Aito-usage was slightly simpler than for the other teams, and we couldn’t try it out live, which finally pushed them to third place. We did extend our price categories to third place as well, though, since we felt they really deserved something for all the effort they’ve put into this.

The Teams HastyTastyBot and GRAB-A-SEAT were also superb contributions, but we needed to choose one or the other, and they, unfortunately, were just outside the top 3. Great work, though, and in the video production categories, both of these would’ve been on top.

Key takeaway - the attitude

So, what’s the big takeaway from this? It was a really intense and rewarding experience. The toll of it is probably clear from my somewhat random rambling here above. Working in a company, even a startup like us, is much more about planning and execution of set goals. We tend to make things “ready” since redoing might be costly or at least cost more down the line. It was a real refresher to see that writing software can also be chaotic and all over the place, and still lead to as good results like this. I think it might probably be a good thing to shake the snowglobe like this every now and then to spark new ideas in the process.

This was a good experience and I’d guess we’re going to have a new go at it again sometimes later. But first, a week of summer vacation.