This was the problem statement we were tasked with on December 3rd, 2023 for our first student consulting challenge. We had two weeks to solve it.
Here, I'll share our final recommendation deck, an overview video, as well as some lessons I learned from this experience.
I started off on the wrong foot. After dealing with low morals as a result of recruiting three people who weren't going to do anything, I kicked them out a week in. I joined another team because I wasn't going to waste any more time dealing with people with no motivation.
The new team I joined was much better. One of the standards this new team held was that communication was key. And communicate we did: we had meetings every day, 15 minutes minimum. Often it would extend to 30 minutes, and the last sprint was a three-hour grind.
Although it was tempting to jump in with the first "good" idea that pops into my mind, there was no research to back up any part of it in the first place. Starting with research helped us understand what exactly Google wanted from us.
With research comes a mess. With so many sources, it would have been impossible to keep track of all of them, their significance, and not end up with 40 pages of illegible descriptions. Organization was crucial to make sure that we could come up with conclusions from what we found.
It was satisfying to finish the deck and overview video at the last minute right before the deadline. Still, the quality of our work could have benefitted from having more properly managed time, which was a result of starting in a bad team and overestimating how long two weeks were.
... how can we sustainably continue the accelerated AI growth while saving our planet and optimizing costs? That's what we wanted to find out this time. With lessons learned from the previous challenge, I was ready to take on this heavily technical 4-week-long challenge.
This time, I chose the right people. At least, that's what I thought, and off I went to a week-long school trip. I came back, and it took another week to realize that things were off. Again, I had to kick people out, and for the second half of the challenge, me and only one other person persevered through it all to make our recommendation possible.
Don't get me wrong. Working on this with only two people in the team is hard, and doing this as I give up my spring "break" is harder. We were working despite time zones, travel, and bedtime curfews. But it was that positive attitude of "Hey, if we try hard enough this can be a killer of a recommendation" that propelled all of this.
It's important to maintain good relationships, and one way NOT to do that is to give excuses as to why you haven't completed a task. The integrity of the team had broken down by the second week, and the end product could have used some additional brainpower if better leadership had been taken.
Yes, this is obvious. You won't believe how many don't adhere to this though, and spend all of their time making a single slide look "nicer." In the end, we spent at most five hours on design, and the other fifty hours (each!) were on content.
Sometimes, ideas can come up that seem revolutionary. Just do a sanity check and you'll be good. I mean a good sanity check: does the idea really pay off, or is it just "cool"? We ended up sticking to liquid immersion cooling even though it was only "cool" at the start because we found all the evidence we needed to show that LIC genuinely was a great solution.
Good news: our team of two (when the normal team size is around four to five and with half the time) made the Microsoft Challenge top 5 finals. For context, this is out of at least 50 teams worldwide, which was great. However, it was more about getting into finals that we felt good about, not actually performing. All the work was essentially done, so the fun part was already over.
But by the time finals actually came around, it was June. Neither of us on the team cared enough to put in more energy, such as creating a separate presentation deck suitable for pitching to a Microsoft director or carving out the first block of school to present live instead of recording a video. But these action items that we didn't act on weren't the problem. It had just taken too long for the director to get back to us, which meant that everyone cared even less than before.
Shooting for 10x, BlazeWatch was a month-long endeavour to make drone-powered wildfire detection a reality. Check out our final deliverables and the biggest ideas I've taken away from the project.
Presentation Video: https://www.loom.com/share/bb23a290e1ce40459248ce19cb369838?sid=2e5e8e6b-835d-47df-b17a-1a8aa8402bee
90-second Video Pitch: https://www.loom.com/share/68ba9751430441d09bf6e1d45feaa928?sid=a5a63b45-8025-4636-bc53-3fd2914116e7
Website: https://blazewatch.typedream.app/
Article: https://medium.com/@leonardozhou2020/blazewatch-a5c7a6b3154c
May is the month of exams. This meant a good two weeks were dedicated towards APs, not Moonshots... and because everyone's two off weeks were the last two weeks before presentations, it was necessary to cram all of the work in the first two weeks instead. Is it called reverse procrastination?
It's impossible to come up with a good idea from step 1. Some ideas were just silly, and others weren't 10x enough and had already been implemented. Then, ideas started sounding cool, but economically or physically unreasonable. We switched ideas around so much that most of the first two weeks of Moonshots had disappeared. So much for reverse procrastination.
But it paid off. After bouncing budding ideas back and forth, we finally landed on one that made sense, was still hard to implement, but could work to solve a problem in an effective way. The entire trajectory of our Moonshots would've been different if we had chosen a different idea.
Individually, having to juggle APs and working on Moonshots would've been impossible. However, just by setting the standard on getting work done in meetings, we could all contribute our share of energy to complete the Moonshots with a bang.
Needless to say, we had to prioritize getting the deliverables done (all of them!) over making them high-quality due to our limited amount of time. Although quality could've been higher for some of our deliverables (website, article), we still got it done.
We also got into finals for the Moonshots (we beat six other teams in our cohort, and made it to the top 11 teams spanning the globe because our idea was realistic and useful, not to mention cool). This one mattered quite a bit more to our team than the Microsoft finals did.
At Moonshot finals, there was one major flaw that almost everyone had (except for the winner, who bettered everyone else in almost every way): pitching the idea like it was Shark Tank. Especially when presenting a crazy idea to a bunch of smart people who know what technologies you are planning to use, it doesn't make sense to try to pitch the idea. Unfortunately, it's quite natural to do so.
TKS started in September, and to the outside observer it's winding to a close in June. But to the active participant, TKS really started at different times for all of us (for me it started in January), and it never ends. A presentation at 10 months is only a landmark to show how much we've gone - and how much more we'll go.
In this presentation, I talk about how I had forgotten how to build with my head buried in the academia game, but rekindled that love through projects that forced me to go in head-first and subsequently taught me much more than I had expected.
In my first hackathon, I collaborated on DaVinci Solve, a LeetCode interview simulator that encourages users to talk through their thought process. We combined APIs from Groq, OpenAI, Leetscrape, Beautiful Soup, and Gradio to end with an MVP in 36 hours.
Check out the repository here: https://github.com/luckyducky037/DaVinci-Solve
Copyright © 2024 leonardo-zhou.com - All Rights Reserved.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.