A jack of all trades is a master of none, but oftentimes better than a master of one.
Hello, I'm Brett. Computer scientist by background, but among the cadre of life-long learners, with my passions taking me in a variety of directions. I began coding on the TI-83 calculator, then for an aerospace company at 17, helping to optimize their supply chain. I moved into the hedge fund world during university, went on to work at the biggest fund in the world (at the time), then Forrest Gump'd my way into building a systematic trading fund from the ground up in London, eventually managing 150 million in AUM. After that I decided to port my skills as a data engineer into the tech startup world, helping small companies with lots of data figure out how to use it.
Troubadour Technology
My independent data and cloud consulting practice, architecting data platforms, AWS infrastructure, and analytics for clients with significant data needs. See the roles below and Projects for further detail.
SNEWPapers
A self-funded passion project (live at snewpapers.com) applying document-layout analysis and frontier LLMs to historical newspapers at a scale and depth not done before. Roughly 2,500 hours in over the last year, with substantial self-funded infrastructure cost.
Built solo end-to-end: gathered a million images from the Library of Congress, investigated frontier document-analysis tools, OCR tools, and LLM models, then customized them to suit my purpose, as well as training custom vision models for specific processing steps.
Built an AWS architecture with CDK to store images, pre-process them with a fleet of GPUs, and orchestrate post-processing jobs with Airflow.
Custom debugging tools and metrics suites for validations.
Dozens of prompt-engineering scripts working together for post-processing and categorization, as well as infrastructure tools for managing and monitoring batch processing jobs.
Designed agentic Grok “Sleuth” assistants with tool-calling that stream live progress over Solid Queue, Valkey, and ActionCable WebSockets; built a custom Lua/Valkey token-bucket rate limiter to gate AI usage by subscription tier.
Indexes 6 million stories from 3,000+ American newspaper titles spanning the 1730s through the 1960s, with hybrid BM25 + FAISS HNSW kNN semantic search and faceted filtering.
DetailPage
Built and ran the entire data backend for DetailPage’s Amazon marketplace analytics product: a double-digit TB scale Amazon Redshift platform ingesting hundreds of millions of products a month. Built it solo, then onboarded and trained a small team of engineers to take over the project.
Kept the entire platform running on under $3K/month of AWS through serverless, auto-scaling, right-sized infrastructure, all defined as code with the AWS CDK.
Designed a 24/7 distributed ingestion pipeline (SQS + Lambda fan-out, token-aware rate throttling, priority tiers) pulling product, price, and search data from four external vendors and running dozens of orchestrated ETLs.
Built ML pipelines forecasting product sales by combining scraped marketplace metrics, customer actuals, and Amazon’s self-reported “soft ranges”.
Designed custom category-universe creation for enterprise clients (tailored taxonomies built when no Amazon category fit their lineup), unlocking reporting and competitor analysis unavailable anywhere else.
Delivered the FastAPI services powering DetailPage’s SEO and AEO optimizations, on top of a Redshift warehouse covering tens of thousands of product categories and millions of keywords across dozens of client brands.
Literacy Pro / Pairin
Rebuilt the company’s entire data platform, replacing a costly, siloed Pentaho ETL architecture with a serverless Python platform on AWS, designed and built solo. Worked directly with external data vendors to standardize incoming data across the integration.
Cut annual AWS cost from roughly $250K to $10K (~25×) and reduced data processing time from days to minutes.
Built a normalized 130+ table data vault and a custom record-linkage engine matching millions of adult-education participants across 10+ siloed source systems on 15+ identity signals, with a full audit trail of every match.
Designed ingestion from a dozen-plus external systems (state reporting systems, community-college exports, workforce-agency APIs, SFTP/FTPS feeds), landed through AWS Transfer Family and S3 and orchestrated by Airflow/MWAA on ECS Fargate.
Restored reporting accuracy across hundreds of locations; delivered Tableau dashboards and secure partner APIs, with the full platform defined as code (AWS CDK).
Macrovesting
Macrovesting was the natural evolution of my work at Sequent Capital. While at Sequent, a large part of my job was macroeconomic research and designing algorithmic trading systems. On this aspect of the business I worked alone, and I had a list of ideas longer than I had time to pursue them. I sought out a community of algorithmic traders to get ideas from and share my ideas with, but it didn’t exist in the form I was seeking. So, I set out to build a platform that:
Was a collaborative community based around systematic trading for quantitative traders and enthusiasts.
Provided traders with the ability to design systematic strategies in the equities, futures, and FX spaces, without having to worry about the tedious processes of gathering and managing data.
Provided educational support, pragmatic interactive examples, and a collaborative community focused learning environment, where users are able to get up to speed quickly, create their own tradeable strategies and portfolios.
Had services to construct strategies, visualize their performance in a variety of ways, construct portfolios from collections of strategies, analyze risk metrics and historical attributions, optimize at the strategy and portfolio level, and most importantly, create tradeable systematic portfolios.
Validated by replicating the Salient Partners $14B Risk Parity Index with ~88% correlation across a 26-year backtest.
Sequent Capital, LLC · London, UK
I was the second employee at Sequent Capital, LLC, a subsidiary and the quantitative trading arm of the Fund of Funds, Bainbridge Partners, LLC in London, UK. My prior work at Highbridge Capital in Risk Management was helpful to me, as well as having completed the Chartered Alternative Investment Analyst (CAIA) program, which focuses heavily on Quantitative Trading, however there was still a steep learning curve! I was initially hired as a developer, to design the architecture and code to enable the fund to trade a blended Risk Parity and Global Macro Trend product, focused on global equity, fixed income, currency, commodity and volatility assets. In the beginning we were very much in startup mode, and in my capacity as developer, along with my colleague, I worked seven days a week for the first two years in order to get our initial product off the ground. The things that needed to get done first were:
Build databases (MySQL) on AWS to house historical Futures, FX forward and Equities data, as well as general market and economic data.
Build accurate historical time-series (back-adjusts) of our historical data, taking into account trading costs and assumptions. Specifically, slippage assumptions, trading fees, corporate actions, splits and dividends, futures rolling logic and interest rate curve interpolations for FX Forwards.
Build trading strategy libraries to handle entry and exit logic, market patterns, position sizing, incorporating our back-adjusts and extraneous data, handling order types and market / exchange specific considerations.
Designing a robust portfolio construction process. We took a modularized approach (think of the portfolio as a tree structure), where we could handle risk in a variety of ways at the strategy (position), asset class, sector and portfolio levels in different ways.
Implementing stringent risk management policies. Similar to the portfolio construction process, I enabled risk management procedures at the strategy, asset class, sector and portfolio levels. The goal was to monitor overexposure at any of these levels (and ideally not to ever let it happen to begin with).
Create a separate system for the automated trading, one which would monitor our holdings, our signals, and execute automatically, as well as integrate the flows into our back-office.
Create a front-end to visualize our holdings, strategies, performance metrics, PnL attributions, risk levels etc.
Developing a research process for gathering data, creating/improving/optimizing trading strategies. Amongst many considerations, this involved figuring out what data was meaningful, how to combine price information with exogenous market data, how to avoid overfitting, how to determine reliable objective functions and how to monitor ongoing performance.
Monitoring our automated processes, making sure our orders for the day were in our automated platform and monitoring them throughout the day, as well as dealing with rebalancing of the portfolios due to investor in/out-flows and risk based rebalancings.
Developing an optimization process, originally brute force, but eventually genetic, leveraging EC2 on AWS to distribute the work quickly.
Over the next four years, the fund grew from 10 million in assets to 150 million, and from one portfolio to three portfolios.
Diversifying our initial stable of 20+ medium-term trend based strategies, into 60+ short and medium term strategies. The new strategies involved mean reversion, relative value, fundamental filters and rankings, asset specific and inter/intra commodity spread based strategies.
The results of my efforts to improve the trading systems and portfolio processes resulted in a net gain of approximately 20% in realized returns vs. the returns we would have achieved had we retained the original trading strategies. This was achieved during what was arguably one of the most difficult times in decades for funds in the Risk Parity and Global Macro Trend spaces.
JP Morgan Chase · NYC
Working on the swaps flow and structured products trading floor developing quantitative trading solutions, enabling traders to price products more effectively and increase flow.
Highbridge Capital · NYC
This was my first full-time role after University. At the time, Highbridge was the largest hedge fund in the world, in terms of assets under management (AUM approx $30 Billion). My primary role was to support the Chief Risk Officer (CRO), and his team of six. However, I also developed solutions for the Mezzanine debt group, Quantitative Portfolio Construction (QPC) group, various Long/Short Equity (Asia, US, UK) PMs, Statistical Arbitrage, Global Macro, and Options PMs. Some responsibilities were:
Developing, maintaining and improving risk management tools (mainly Java / VBA / SQL) for analyzing risks at the PM level, as well as the fund level.
Developing and maintaining over 100 interactive applications, automated reports, numerous trade blotters, position and activity applications, P&L, exposure, parity and allocation reports for the abovementioned groups, enabling the stakeholders (mainly Risk, PMs and Traders) to focus less on mundane tasks and more on their core expertise.
Spearheaded the migration of the Risk Group’s existing legacy tools into a more modern and scalable solution, decreasing the time on myriad daily automated tasks significantly.
Worked extensively with the head of the Quantitative Portfolio Construction (QPC) group utilizing Barra Equity Models to break PM P&L into some 70 Style, Industry and Country factors, allowing PMs to better understand the areas they provided alpha in and deploy their risk capital more efficiently.
Aptus Capital · Virginia, USA
My first hedge fund role. Developed tools (VBA, MySQL) for the traders to collaborate on ideas.
Smiths Aerospace · Maryland, USA
My first programming job (VBA, Oracle). Worked on solutions to help optimize their supply chain, while they were going through a process of reducing suppliers from 100+ to 10+ core suppliers. I had to teach myself VBA and SQL on the job.
Imperial College London
A 3-course specialization covering linear algebra, multivariate calculus, eigenvalues/eigenvectors, and PCA.
Columbia University
A 3-month, 12-module specialization, heavy on the math behind common algorithms: regression, Bayesian methods, classification, supervised and unsupervised learning, clustering, recommendation systems, sequential models, association analysis, and model selection. Uses scikit-learn.
deeplearning.ai
A 5-course specialization on the math behind, and raw implementation of, deep learning and neural networks. Uses TensorFlow.
fast.ai
Courses using the fast.ai library on PyTorch, focused on image classification, SGD, NLP, and CNNs, with heavy emphasis on data preprocessing, feature engineering, and performance tuning.