My role at Autumn Compass was about hunting for profitable strategies in SSFs, forex, equites, options, crypto and ETFs as well as for improvements to AC’s existing algorithms. I got a lot of freedom to investigate what I judged to be promising leads and I've returned several robust discoveries.This involved hypothesising market inefficiencies, distilling mathematical models that could exploit those inefficiencies, coding those models (usually in python) to test against historical data, cleaning that data and marshalling statistics to understand the results. This analysis has used Bayesian statistics, computer simulations, and where necessary, a few statistical methods of my own invention.Autumn Compass' trading strategies are produced with the help of AI. Through discussing, testing, implementing and debugging improvements to these systems I have developed a deep understanding of the dynamics within the common AI models. In 2020 the company received $500,000 AUD in government funding to study AI and my work was the majority of the company's research submission at the end of that year.At Autumn Compass my software skills have improved massively. After 4 years of coding every day, I have become highly proficient in python as well as SQL, C++, the Linux operating system and collaborating with other developers through GitLab.

Created and managed a database of teaching resources.Mentored other coaches to become great teachers as well as great debaters.

This experience has developed my problem solving skills, requiring me to target lesson plans toward the needs of individual students and my ability to communicate complex ideas (in philosophy, economics, politics and oratory) in an understandable way.I have had the pleasure of coaching 84 young debaters across 3 classes and 5 teams.

Inspired by an episode of the Planet Money podcast about early quant-funds I began to build my own trading algorithm.Earlier versions of the algorithm predict share prices in the short term using recent price movements, historical volatility, systemic risk and market capitalisation. This happens, using a non-linear regression model, with a flexible functional form. The idea was back-tested against 10 years of Bloomberg data from Australia, the USA, Canada, UK, EU, Japan, Korea, India, Switzerland, Russia, Brazil and Iran. About a year into the work, the project got labelled Suvorov Securities after the diligent logistical obsession that served Russian general Alexander Suvorov so well.The current version (v 71.2) uses the current prices of options to create a set of probability distributions for the underlying stock price over time. The moments of these distributions map them into 4-dimensional space where my own statistical method derives the actual distribution. This second distribution is used to analyse the payoff from purchasing both stocks and options. The strategy has been tested in ASX markets with an annual return of 15.2%.Skills acquired:• Bloomberg dexterity (terminal and APIs)• Coding in R, Matlab and Python• Granular understanding of transaction costs• Knowledge of options and equity marketsAdditionally, I participated with this project in the University of Sydney’s Genesis Program, to develop fledgling start-ups.