• What is Tshong AI about? What did you personally contribute to this project?

Tshong AI was born from an idea I came up with alongside a friend while I was on a sabbatical in Bhutan. We noticed that many shops were snapping photos of payment receipts with their phones and recording journal numbers manually. We wanted to simplify this process and automate the whole thing. The goal was to cut down time from hours to just a few minutes. Thanks to a handful of supportive customers, we were able to keep improving the system based on how their staff interacted with the platform. I was involved every step of the way, from brainstorming ideas to building, iterating, and working directly with customers.

• What technologies or tools did you get to use during those projects? Anything you enjoyed working with?

I worked with vanilla JavaScript and Node.js, integrating tools like Google Cloud Vision API, Railway for deployment, AWS S3 for storage, and DigitalOcean droplets for hosting. Since the product was constantly evolving, we had to switch tools as we learned more about what our customers needed, so I got to experience a lot of different technologies along the way.

• Were there any interesting challenges you faced while working on those projects? How did you go about tackling them?

One of the biggest lessons I learned was that just because you build a new feature doesn’t mean users will love it or use it the way you expect. When I built the first prototype, I thought it would be enough, but I quickly realized the real work was in paying attention to how users interacted with it. We had to tweak things based on feedback, and sometimes users showed us what was really important. For example, I didn’t realize the system needed fraud detection until a customer pointed out that people were paying but only getting screenshots, not real transactions.

• How was the experience of collaborating with your team members? Any memorable moments or lessons learned?

Working with my team and customers was incredibly rewarding. One moment that stands out is when I was chatting with a customer and offered to help them use our system for tallying. In some Asian cultures, communication is more indirect, so they agreed but asked us to come back another day. One of my team members pointed out that they might not want to share sensitive sales data with us. I realized they were right, and we decided not to push them. It was a great reminder to respect cultural nuances in business.

• What were the outcomes or results of those projects? Anything you're particularly proud of or found surprising?

What I’m most proud of is seeing the real happiness from our customers when they used our system and saw the fun party animation pop up. It felt great to know they were genuinely enjoying the product. I’m also really proud of the assistive capture system we developed, which helped users take the perfect photo at the right angle. This feature helped boost user engagement, and that made all the hard work worth it.

• When it comes to writing code, do you have any personal approaches or strategies you find effective?

Since we were moving from design to product creation, I focused on getting things done quickly with the help of tools like Cursor AI. Instead of obsessing over best practices right away, I gathered data from customers based on how they were using the product or what problems they were encountering. Then, we’d make updates based on that feedback. After a few weeks of use, we’d optimize the code for best practices once we were sure the feature was genuinely needed.

• Can you think of a specific problem or task that was particularly challenging, and how you approached solving it?

One challenge we faced was when customers started taking receipts so quickly that our AI couldn’t read them accurately. To solve this, we built an assistive capture system that recognized when a phone was being used, triggered a 3-second countdown, and guided the user to take a clearer photo. This improved accuracy from under 50% to over 96%, which was a huge win.

• Were you involved in any aspects of project planning or management? How did you handle deadlines and prioritize tasks?

Since this was a stealth startup, I worked closely with my advisor to review our team’s progress and the product every week. We found that project planning is important, but it’s the people behind the project that really make things happen. We focused on getting the right people in place and making sure they had the support they needed.

• Looking back, what have you learned or gained from those experiences? Any skills or insights that have stuck with you?

One of the biggest insights I gained from building this startup was that earning customer trust is far more important than having the most advanced technology. A prototype is essential, but customers will tell you how they want to use your product. I learned that "just works" is often more important than "perfect" when it comes to product development. You can always refine a feature after you’ve validated that people want and need it.

• Is there anything you'd do differently now, based on what you've learned from those projects? Any feedback or advice you've taken to heart?

The biggest thing I’d do differently now is to assess infrastructure more carefully before scaling a product. Our system required customers to have at least 10 Mbps internet, but we discovered most of them had only 1 Mbps. We couldn’t improve the infrastructure without a massive overhaul, which ended up slowing our progress. It taught me that technical constraints can really impact how smoothly a product scales.