About a month ago I went for HacX 2025, which was a competition that was organised by Microsoft and HTX in collaboration in SUTD. I had a lot of fun during this competition, and I wanted to reflect on what we did and how we did it.

What the HTX competition was all about

The overarching team for HacX was public safety, where we had to innovate and design solutions that have a positive impact on public safety. These were among some of the problems statements that we could choose from for the competition:

1. Future of the Neighbourhood Police Post (NPP) / Neighbourhood Police Centre (NPC) Service Zone
2. Future of the SOD's Operational Gear and Equipment
3. Future of the Prison Transporter Vehicles
4. Future of the SPF's Driving and Riding Training

Our team decided to choose the first problem statement, which was about Future of the Neighbourhood Police Post (NPP) / Neighbourhood Police Centre (NPC) Service Zones

What we had to do

For this statement, we had to design 2 main components of a system for NPPs and NPCs:

1. The Self-Help Kiosk: This kiosk would be where those that need assistance would go to to enter things like lost items, or report lodging among others. These kiosks would be unmanned, which means that they would need to work autonomously with no input from a police officer.
2. The dropoff system: The kiosk allows for the depositing of lost items, as such a dropoff system also had to be implemented to allow finders of lost items to deposit them for safekeeping.

What we did

We decided to build off the starting point that current automated kiosks in NPPs and NPCs had, since it was a pretty good foundation to start building atop off. Then, we overhauled the UI of the kiosk to give it a much better look, and also to optimise it for a new agentic workflow we would be installing into the application later.

Agentic workflow?

As part of the overhaul, we decided to link the form up to an AI persona called Officer Clif. Officer Clif would allow users of the kiosk to describe the situation in words, and automatically fill up the form on their behalf. It would ensure that the user affirms what Officer Clif enters into the form, for accuracy. Should Officer Clif be unable to transcribe what the user is saying for any reason, the UI still allows users to manually enter the details, ensuring that the kiosk stays usable even if Officer Clif is unable to assist.

To achieve this, we exposed the form as a tool for the AI agent, which allowed it to manipulate the contents of the form, while also allowing the user to confirm or deny changes that Officer Clif makes.

How about the dropoff system?

My team designed a modular dropoff system, that could be adapted to the different sizes of NPPs and NPCs, allowing for the space used to be optimised. It was controlled with a arduino which was in turn controlled with a serial interface to actuate the motor that open and closed the doors. Though, the main highlight of the product definitely had to be the Agentic Capabilities of the new UI.

The presentation

Although we were all discussing 5 minutes before our turn what we needed to do, in the end we managed to present quite well, and we even managed to clinch third place in the category! It definitely came as a surprise to all of us. Unfortunately only our team leader was there to collect the prize as everyone else was gone due to other reasons.

It definitely would not have been possible to achieve this with my team.

What I took back at the end

From this, I had learnt how to properly integrate an AI solution into a website to unlock some truly novel capabilities and new ways of interaction for the end user, instead of just sticking a chatbot to the bottom-left or right hand side of the page and hoping that people actually use it.

— Sairam