Our client, founded by former leadership from Bridgewater, is a FinTech company on a mission to democratize access to alternative investments. Leveraging ETFs (Exchange-Traded Funds), their unique offering empowers the average investor to “invest like the 1%” without paying the notoriously high hedge fund management fees. Fueled by the successful launch of their first ETF product, their ambitions grew to include launching several new ETFs, each of which aims to replicate the returns of other well-known but uncorrelated types of hedge fund strategies.
However, the operational complexity required to implement and execute this deeply technical, machine learning-based approach — including the manual orchestration of an ML pipeline — laid bare the technical debt that needed to be addressed before the business could scale. The team recognized the need to harden their infrastructure and automate the execution of the application logic; a failure to do so could lead to model drift or untenable slippage to the benchmark. Their original approach, aside from being labor-intensive and making extensive use of spreadsheets, required meticulous double-checking and reconciliation steps to ensure no errors were introduced with each model run.
In order to unlock the path to growth, the imperative was clear:
- Establish a modern, robust cloud infrastructure
- Introduce a layer of automation
- Harden the workflow to increase resilience
APrime Technology designed and implemented a simple, centralized yet scalable infrastructure, fostering improved day-to-day operations, fortifying systems crucial to the company’s mission, and establishing a foundation for product expansion.
First, we streamlined and consolidated the disparate set of business-critical, manually-run algorithms and scripts into a single data pipeline managed by automated deployments. Crucially, we incorporated familiar and approachable technologies into the modernized setup, producing the desired gains in efficiency without overwhelming the existing team with a completely new tech stack to ramp up on.
We set up the data pipeline on AWS Sagemaker allowing for repeatable model runs, elastic and scalable compute, and more readily replicable development and execution environments for the growing team. Github Actions were used to create automated deployments for the rapidly growing code base.
In parallel, we migrated the backend databases from a local, outdated SQLite setup to a Postgres Instance in AWS RDS. This migration not only modernized the infrastructure but also optimized database management and accessibility.
In three short months, APrime Technology upgraded and migrated the customer’s infrastructure, streamlined the application logic, and automated core components of the ML pipeline. These efforts laid the groundwork necessary to proceed with the launch of several new ETF products, leading to a substantial increase in AUM while freeing up time for the co-founder and Chief Data Scientist to focus on their core value proposition — the statistical research required to develop new investment strategies.