This Data Scientist Knows Compliance

Hi. I'm Jeremy Doyle, a data scientist with extensive regulatory compliance experience in the securities industry, I’ve dedicated my career to helping ensure investors receive fair and equitable treatment and are protected from fraud and abuse. Firms that fail to shield investors from impropriety face significant regulatory and financial risks. Leveraging data and technology to manage these risks at scale has never been more important. I’ve therefore developed expertise in machine learning, artificial intelligence, probability, statistics, data analysis, and software engineering to apply these disciplines to compliance risk problems.

Skills & Expertise

Broker–Dealer Compliance
Investment Adviser Compliance
Financial Crimes Compliance
Machine Learning
Statistics/Probability
Data Analysis

Python
SQL
HTML
CSS
JavaScript
Agile Development

Jeremy Doyle's expertise lies in both Regulatory Compliance and Data Science.

Data Science Projects

Most of my data science products are the result of professional work and, therefore, cannot be shared. However, at a high level, some recent professional projects include:

An NLP classifier that categorizes unstructured memos into: 1) memos that are important for the task-at-hand, 2) memos that can be disregarded by the task, or 3) memos that instead require human classification.

A model that selects high-risk financial advisors for in-depth sales practice investigations using an ensemble of three anomaly detection algorithms (Isolation Forest, Histogram-Based Outlier Detector, Copula-Based Outlier Detector).

An anomaly detection model used to identify Structured Products purchases that may be unsuitable given a customer's characteristics and financial circumstances.

A bespoke algorithm (which serves as the basis for multiple fraud detection scenarios) that matches two strings based on the named entity variations found within them. For example, "Everstay Funds FBO Joseph and Sally Peterson 123 Main St" matches "Joey Peterson 4321 Elm St".

Since I am unable to share the details of this work publicly, I've completed several personal projects with public data — summarized below — that demonstrate the quality of work I am able to produce.

SimCAP Python Package

Overview

SimCAP was created with the intent to make it simple to generate useful simulations of correlated multivariate financial time series. With SimCAP, users can simply provide a pandas DataFrame of historical stock prices (or any other financial instrument) and easily generate thousands of simulations in minutes. Simulations resemble the original time series in both the correlations between assets and the statistical properties of the returns distributions.

Model Type

Hidden Markov Model

Training Data

Users provide a SimCAP instance a pandas DataFrame of historical asset prices. While example datasets are provided with the package so users can quickly demo functionality, the data needed to train a model is provided by the user.

Intended Use

Possible uses for SimCAP are trading strategy development, portfolio optimization, portfolio risk management, financial planning, augementation of data sets for machine learning, etc. Easily installed with pip.

pip install simcap

GitHub Repo

RIA Similarity App

Overview: The RIA Similarity App is a recommendation system (hosted on Heroku) that allows users to select 1) a Registered Investment Adviser (RIA) and 2) a region of the US (or all regions) to be shown up to 20 of the most similar RIAs to the target RIA in the selected region.
Model Type: Unsupervised k-Nearest Neighbors
Training Data: The models used by the application are fit with data obtained from US Securities and Exchange Commission (SEC) Form ADV. The data selected for the models come from Part 1 of the form which captures structured information about each RIA's business, ownership, clients, business practices, and affiliations.
Intended Use: The application presents similar RIAs to a user's target RIA based on multiple attributes related to business model, scale, and growth. For a user responsible for marketing to RIAs, recruiting advisors from RIAs, or acquiring RIAs; this application could aid in the discovery of firms that are like an RIA the user has had success with in the past — helping the user focus on prospects that might be receptive to the user's value proposition given the similarities to the target RIA.
: Open App GitHub Repo

Licenses & Certifications

Organization	Credential
FINRA	Series 4 — Registered Options Principal
FINRA	Series 7 — General Securities Representative
FINRA	Series 24 — General Securities Principal
FINRA	Series 27 — Financial and Operations Principal
FINRA	Series 53 — Municipal Securities Principal
FINRA	Series 66 — Uniform Combined State Law
Scaled Agile	Certified SAFe^® 5 Practitioner