Brian Lee Yung Rowe is the face of Zato Novo. A 21st century polymath, Brian draws from a deep well of experience that spans startups, quantitative finance, product development, and software engineering.
Formally educated in mathematics and electrical engineering, Brian has leveraged the scientific method throughout his career. Having spent six years teaching graduate courses in natural language processing, machine learning, and big data, Brian applied his teaching experience to develop his Teaching as Management (TAM) method. TAM combines self-determination theory (SDT), social and emotional learning (SEL), the scientific method, and lean production to create a sustainable data-driven culture that minimizes management overhead.
Models require data, but what can you do if there isn't enough data? When asked to design a credit model for a new loan product, we leveraged our experience with idiosyncratic time series models to create a tunable credit model. Doubling as a simulation engine, the lender can stress test the model against numerous macroeconomic scenarios.
In liquid markets, we expect the market to tell us the price of an asset. In illiquid markets, it's not as straightforward. When asked to build out a quant group at a bond exchange to provide fair value prices for a universe of 22,000 bonds, some were skeptical. How can a small team of 5 accomplish in six months what a group of 30 PhDs couldn't achieve in 3 years? Six months later, our real-time pricing model bested the industry benchmark by 30%.
A multi-billion dollar discretionary hedge fund wanted to expand into quantitative trading. We were asked to build a quant team and create an initial research platform. Working closely with the Chief Economist, we built a graph-based system to easily define time series based on underlying price data alongside fundamental data and macroeconomic datasets.
An aggregator of peer-to-peer loans wanted to introduce an automated investing product. We created a bespoke credit model to predict probability of default and loss given default (LGD) for new loans. Loans were partitioned into tranches and allocated to investors based on their risk profile.
A legal startup wanted to automate contract review by comparing a contract with a gold standard. We created a machine learning model to classify clauses within a contract and identify the differences between it and a gold standard.
Budgeting is a tedious exercise. What if a model could predict your spending and predict your expenses for the following month? Using just three months of bank transactions, we created an idiosyncratic time series model to predict the expenses for individual consumers to within 94% accuracy.