Yonder is the coolest startup in the UK, "the Monzo of reward credit cards", and the "Best Newcomer" according to British Bank Awards. I joined Yonder in 2022 as its first Data Scientist, employee #17. Here's what I do at Yonder:

Machine Learning: I use Machine Learning to predict people's behaviours: How likely is a user going to miss a payment? How likely is a user going to upgrade their product? I build these prediction engines into Python APIs, so that Yonder can get live predictions to make personalised lending and product decisions. 

Recommender System: I use Language Models - the tech underlying AI bots like ChatGPT - to rank rewards based on user behaviours. This makes the recommendations nuanced and intuitive. Again, I deployed it as a Python API, so that Yonder can personalise rewards for users in-APP.

Social Listening: I help analyse the conversations generated by Yonder's social media content. I use Natural Language Processing to discover topic clusters and positive engagements, and I use Network Analysis to discover interest communities - this is our cool new project where tech joins creativity!

Team Algorithms: Yonder is a startup that does a lot for team dynamics, such as arranging peer-to-peer feedbacks and small-group team meals. I work with Yonder's People Op. and write mathematical algorithms that help make these team activities as effective as possible.

I am fortunate to have continued academic work with my mentors in my spare time, now collaborating with researchers from Cambridge Psychology, Cambridge Public Health, New York University, King's College London, and University College London. Here are my current projects:

Artificial Societies: We got a bunch of AI chatbots to chat with each other, and made an Artificial Society. Turns out they mimics aspects of real human societies as well. We are now working out exactly how we should use artificial societies, especially what they can and cannot do. Hopefully, we'll be able to simulate A/B tests on artificial societies before testing on real human societies, and enhance current Agent-Based Models in economic and evolutionary research.

Artificial Personalities: As I built artificial societies, I found the chatbots painfully generic. Can we get AI chatbots to mimic diverse human personalities? I am experimenting exactly this: directly "tuning" the weights inside the AI models, to get them varying along human personality dimensions. Relatedly, I am planning to explore using social evolution to fine-tune AIs within artificial societies, mimicking how we humans evolved socially, in the hope to create socially-aligned AI systems.

Online Social Divisions: Our societies today are facing growing social divisions, and social media might have played a role. I am continuing my work with researchers at Cambridge and NYU to study online social dynamics, by using Natural Language Processing to study people's discussions, and using Network Analysis to study the shapes of communities. The goal is to find ways to maintain open debates but bridge social divisions.

When I got my first paycheque (thanks Yonder), I pledged to donate 10% of my disposable income to support interdisciplinary explorations:

The James K. He Scholarship is dedicated to support data-driven, interdisciplinary research at the undergraduate level. Open to students of ALL subjects at Selwyn College Cambridge, with Preference given to projects from disciplines with limited fundings.

Why did a 21-year-old set up a scholarship?

In most of human history, innovators are those who love building bridges, those who refused to be defined by a single way of thinking. Sadly, there is hardly any funding in the UK that supports innovative, cross-discipline research, especially for undergrads. 

So, I decided to put money where it counts - I've enjoyed the self-sovereignty of not being defined by a single discipline, and I hope to build a community of innovators who refuse to be defined.