My thesis studies how risk aversion affects the behaviour of learning agents in a financial market simulation. The setting is a limit order book, where orders to buy and sell are submitted, cancelled, and matched continuously. I focus on two common trading tasks: market making and order execution.

Most sentiment‑based market analysis focuses on the average mood. But averages can hide conflict: a day where half the posts are very positive and half are very negative can average out to “neutral,” even though the market conversation is highly polarized. In this analysis, I focus on sentiment disagreement, i.e. how divided sentiment is, rather than just the mean.

A few weeks ago, I participated in the WTC 40 Hour Hackathon – AI Builders, centered around a question: How can AI augment our Natural Intelligence?

This project tried to investigate how effectively a centralised takeover strategy can mitigate traffic congestion in simulated highway based on agent-based model. This was done through the reproduction of an already existent agent-based model based on a paper by . The model was expanded upon by adding agent behaviour for multiple lanes and logic that simulates a central controller. It was hypothesized that centralized vehicle takeover will be significantly more effective than human-based models in mitigating congestion. This was confirmed as the central vehicle takeover managed to increase the average speed of the vehicles and maintain a higher flow under congested conditions.