How I (Almost) Won the World's Biggest AI Hackathon...With Zero Coding Experience
Overcoming coding FOMO as a Product Manager.
Background:
When Sam Altman unveiled ChatGPT in 2022, the world errupted in a frenzy of excitement over this “new era” of technology. My friends pursuing careers in software engineering jumped on this new opportunity, embarking on anything from LLM fine tuning to creating their own neural networks.
But even as someone who enjoyed building projects, this new craze of excitement was shortlived. I didn’t know how to code, and ultimately felt lost with projects flying around me at all times. It was the first time I felt coding FOMO. Realizing this complication, I decided to jump into the deep end and enrolled in the world’s biggest AI hackathon to really get a feel of how my skillset could transfer to this new era of projects.
Throughout this article, I’ll take you through the lessons I learned through competing against over 1200 students at UC Berkeley’s AI Hackathon, from ideation strategies to how we ended up as a finalist pitching for a $25,000 VC investment. Whether you’re a software engineer with years of coding experience or someone like me, just trying to navigate through what feels like a code-dominated world, I hope you can draw from my experiences and lessons to develop your own entrepreneurial skillset.
Phase 1: Ideation
Before the hackathon started, my team and I racked up ideas on possible projects to embark on the following day. Brainstorming was done through one of three ways:
Consulting ChatGPT for untapped issues to solve.
Scraping the internet for public datasets and ideating how they could be utilized in machine learning models.
A combination on methods 1 and 2; applying public datasets to untapped (or even indirectly related) issues.
Here’s what I noticed.
Lesson 1: Rewire how you use AI in your workflow.
Oftentimes, consulting ChatGPT for ideas on problem areas to tackle led to repetitive, basic answers that were far from helpful. I soon found that the ideal workflow is to think of innovative problems and solutions youself, but use AI to instead generate any supportive facts needed to validate your solution.
When asking ChatGPT for “innovative solutions to untapped problems in the healthcare sector,” responses would range anywhere from using drones to deliver health samples to developing health wearables (that already exist). Some may say that it is a user’s job to generate a prompt worthy of a quality response, but there comes a point where the return on investment of fine tuning a prompt simply isn’t productive in fast turnaround environments.
Instead, use AI to determine the feasibility of your ideas. I’ve thrown crazy ideas at ChatGPT, from augmented reality text translators to wearable devices to cure period cramps, all in an effort to see if there was legitimate evidence that could support whatever idea I was trying to pursue. I found that instead of spending time coming up with the most elaborate prompts for LLMs to make interesting solutions, my most innovative ideas came from spending that time repeatedly throwing bold ideas at the wall until one is validated.
Eventually, my team and I settled on our idea: an all-in-one health app to optimize daily habits using AI.
Phase 2: Building
Now, with idea in mind, it was time to start building. We marked our territory at the bottom of Berkeley’s MLK Jr. Building and got to work.
Immediately, I layed the foundation for our teams dynamic and delegated tasks to each team member.
Nick (Me): Product Manager
Creating the PRD framework, slide deck, and devpost
Design and create Hi-Fi mockup in Figma
Marcus: Software Engineer
Backend prompt generation and G-Suite integration
Sponsorship prize optimization (Include as many sponsors as possible for eligibility)
Steven: Software Engineer
Computer vision (AI Food Scanner)
Use data to generate AI Health Diagnosis; integrate ElevenLabs
Sean: Data Scientist
Parsed health data from CSV export of Health App
Reformatted data into readable format for LLM
I got to work designing my interactive prototype of the product, drawing low fidelity wireframes and finding UX inspirations from sites like Dribble. I sat their for hours with my team as we designed, coded, and deliberated our ideas for hours on end. But like the ebb and flow of any work sprint, productivity peaked and it wasn’t long until exhaustion started to kick in.
Lesson 2: Hackathons are a marathon, not a sprint.
While this takeaway might raise an eyebrow to some, hear me out. Hackathons are known for their rigorous all-nighters insinuating the notion of a sprint, but that approach isn’t always optimal.
Just a few months before enrolling in this AI Hackathon, I competed in LA Hacks, a 72-hour hackathon where my team and I built a Google Meet bot for real-time meeting translation. However, my team and I tried the stereotypical approach of no sleep and constant grinding, foreshadowing groans of discomfort and stagnation with our project down the line.
Fast forward months later and I found myself facing a similar situation at this AI Hackathon. Up until that point, we had been working for almost 15 hours straight with minimal breaks before productivity started to slow down. How could I foster the most productive atmosphere for the longest period of time?
The answer lied in short, high-quality breaks. Instead of pockets of time wasted on doomscrolling through Instagram, I focused on creating micro breaks with my team to stay productive while being effecient with our time. From anything as simple as a deep conversation on our quick stroll to dinner to a break to pet the llamas, these seemingly 5-10 minute sacrifices actually resulted in massive amounts of productivity gain.
Phase 3: Judging
As the coding period came to a close, my team and I sat back in our chairs, exhausted from the 26 hours of work that had overcome us. But it wasn’t over yet. We had been told that we had mere minutes to prepare before droves of judges would scatter throughout the halls, eager to learn what we had built over the course of the last few days.
Lesson 3: Know your audience.
One of the most highly valued traits in any type of technical job today is communication. But communication is more than just smooth talking and elegant diction. As an avid communicator, I learned that being able to tailor your conversational approach based on your audience results in more interested feedback from listeners. In technical environments like these, good communicators need to recongize their audience and what they are interested in…all in a split second.
Should I explain our tech stack? Do I explain what API’s we used? Do they even care about what an API is?
As the first judge walked to our table, I immediately started to scan them for possible interests. Apple watch on the left hand, Vivo barefoot shoes, exuberant energy. From these observations, I assumed the emotional appeal of our company (aligning heavily to the problems fitness industry) would be more impactful than a dropping a bomb of technical knowledge. On the other hand, another judge came over to our table with a Meta jacket in hand, staring at our code base without asking a single question about our idea. Safe to say, I wasted no time handing them over to our software engineers so he could pick their brain.
While tailoring your pitch on the fly may seem risky, decisions like these can make or break your hackathon placement. The premise of any pitch should be to leave a positive and lasting impact on your judge, and finding ways to tailor your pitch to itch the judges best interests is the best way to do that.
While the hours of judging went relatively smoothly, I soon looked down at my phone and was met with a disturbing message.
*7 missed calls*
Lesson 2: Seize your opportunities and don’t let go.
It had turned out that while we were pitching to our final judges from Intel, I failed to hear the calls announcing our advancement to the next round. Alarmed, I dialed the number on the phone. I was told that we had been selected as Finalists for the AI For Good Award and a $25,000 investment, but they had moved on and replaced us due to our failure to respond.
“It's too late. I think they’ve moved on without your group…You can try coming up to floor 3 to see if there’s still a chance…” - SkyDeck Representative
To me, the words of “I think” and “still a chance” was enough hope for me to grasp on to as I grabbed my group and bolted up the stairs to the judging venue. Out of breath, I looked up at the two SkyDeck Representatives I had called, only to be met with disapproving head shakes signifying our defeat. It was too late.
Or was it?
Overwhelmed by the effort our group had put in over the last 26-hours without a blink of sleep, I tried everything to explain our group’s circumstances to salvage our last shot at this investment. I explained everything, from the Intel judging scenario to our investments we took to even compete at this hackathon. I was met with constant denial before finally…
“…let me see if I can ask.”
A short conversation later, my group and I were let through and had 1 minute to pitch our idea. We didn’t end up winning the $25,000 investment, but to me, fighting and clinching a spot to even pitch on the main stage was a big enough win for me to take from my second major hackathon.
Executive Summary:
It’s easy to get bogged down by the recent surge in AI, especially for someone like me who is recently breaking into the coding scene. I do think coding is an essential piece of knowledge (especially for PMs) that unlock potential for better connections and leadership with SWEs. I myself still take the time to read up on advancements in code and AI in order to best lead and understand what software engineers are thinking. But at the same time, coding isn’t everything. In my recent hackathon, I served as the PM for my group, focusing more on the planning and presentation of our project while overseeing my team’s SWEs throughout the execution process. I learned how to utilize AI in more productive idea generation, fostered productive environments through microbreaks, and battled my way with admin to get a spot on the final pitch panel. These experiences haven’t completely eridacated my coding FOMO, but instead remind me that there are always valuable contributions to make outside of our github repo.
UCLA Hack no sleep lol
Insightful!