Yes, you read that right – you can chat with a cute otter named Kaya to find out anything you want to know about Singapore (my hometown).

As part of my Master’s programme in content strategy at FH Joanneum, we had a course on "Data-driven content businesses", where we explored with our lecturer Michael Leitner how data supports content strategy. 

One of my favourite tasks in this course was crafting a custom prompt or chatbot to solve a problem. I decided to create Kaya, a chatbot with the personality of an otter that helps people learn more about Singapore. In this post, I’ll share why I chose this idea, the key lessons I learnt from the process and some useful resources for those interested in prompt engineering.

Why a chatbot about Singapore (and why an otter?)

As a Singaporean living abroad, I often receive requests for recommendations about my country. I wanted to see if I could create a chatbot that not only provided useful travel and food recommendations but also had personality – making the experience more engaging and fun.

So why an otter? Otters are well-loved in Singapore, often spotted in rivers and reservoirs across the city. They’re smart, curious and full of personality – exactly the kind of traits I wanted my chatbot to have. (Fun fact, Kaya is named after a popular local coconut jam!)

Three key lessons from building Kaya

1. Personality makes a difference

Initially, Kaya was not an otter. While crafting the chatbot’s responses, I realised that simply providing factual answers wasn’t enough. I needed to add personality – a unique tone of voice, local expressions and even the occasional joke – to engage and delight the users.

I enjoyed crafting Kaya's personality. It took some trial and error, but I eventually decided on friendly, playful but knowledgeable to balance personality with usefulness.

2. Details and formatting matter in prompt engineering

One challenge I faced was getting Kaya to provide useful links when making recommendations, so users can explore more. Initially, I included the instruction as part of a paragraph with other details, but I realised Kaya was ignoring the point.

I edited the prompt after this, by reformatting the instruction as a standalone bullet point: "Include links to websites or sources where users can find out more about the destinations, restaurants or events you recommend."

This formatting change made a difference. Kaya started including links where possible, for example:

3. The importance of iteration

During testing, I noticed that some responses felt too generic, while others needed fine-tuning to maintain consistency in personality. I tweaked my initial prompt multiple times to balance accuracy, friendliness and a natural conversational flow.

This process reinforced the importance of iteration – testing, refining and improving over time. Even now, Kaya is still a work in progress – feel free to share any feedback you have!

Try chatting with Kaya!

Kaya is ever eager to help, whether you're planning to visit Singapore, or just want to know more about the city-state! Ask about the best food, hidden gems, Singlish or even fun facts about the country's history.

I’d love to hear what you think – and if you have ideas for making Kaya even better, let me know!

Want to build your own chatbot? Start here

If you’re interested in crafting your own chatbot or exploring prompt engineering, here are some resources I found useful:

• Learn Prompting – A comprehensive introduction to prompt engineering

• OpenAI Prompt Engineering Guide – Official best practices from OpenAI

• Prompting Guide – A curated collection of techniques for optimising prompts

P.S. If you’re curious about other creative chatbots my colleagues built, I recommend these ones:

• Gustl Gurr: A witty pigeon from Graz who knows all the city’s secrets. Look forward to insider tips and urban legends with a playful sense of humour! (He speaks languages other than German too, although he's wittiest in his mother tongue.)

• Spark(y) Hire: An ideation buddy to help you create a job application or optimise your LinkedIn profile.