Jane Street, Goldman, JPM open repos

A viral X thread on May 23 pulled together a set of public code repositories from some of Wall Street’s best-known trading firms and banks. The list included Jane Street’s magic-trace, Goldman Sachs’ GS Quant, J.P. Morgan’s Perspective, Hudson River Trading’s corral and Two Sigma’s flint. None of those projects is new, but the thread put them in one place and recast them as a map of what firms are willing to publish from inside production engineering stacks. The repositories show different slices of trading infrastructure. Jane Street’s project is about low-level tracing, Goldman’s about quantitative tooling, J.P. Morgan’s about streaming analytics and visualization, Hudson River Trading’s about C++ concurrency and Two Sigma’s about large-scale time-series analysis. Together they offer a public look at how firms package internal engineering work for outside use. ### Why did the thread travel? The X post spread because it bundled recognizable names with usable software. In finance, firms such as Jane Street, Goldman Sachs and J.P. Morgan are usually discussed in terms of secrecy, hiring and infrastructure budgets, not GitHub repositories. GitHub pages show the projects are active public artifacts rather than one-off code dumps. Jane Street’s GitHub organization lists 405 repositories, and magic-trace itself was updated within the past two weeks, according to GitHub. (github.com) Hudson River Trading’s public GitHub page lists corral among its repositories, and Goldman Sachs maintains a public GitHub organization as well. ### What does Jane Street’s repository actually do? Jane Street describes magic-trace as a tool that “collects and displays high-resolution traces of what a process is doing.” Its GitHub README says the tool can trace every function call with about 40-nanosecond resolution, with 2% to 10% overhead, and can render a timeline of call stacks going back roughly 10 milliseconds. Those details matter because they place the project in the debugging and performance-analysis layer rather than in trading strategy code. (github.com) The repository shows how Jane Street has chosen to expose a piece of systems tooling that can be useful well beyond finance, while leaving the firm’s execution logic private. ### Why are Goldman Sachs and J.P. Morgan different cases? Goldman Sachs presents GS Quant as a Python toolkit for quantitative finance built on its risk-transfer platform. (github.com) The bank says the toolkit is designed to speed the development of quantitative trading strategies and risk-management solutions, and says its analytics tools are “trusted daily by over a thousand quantitative developers” at Goldman Sachs. J.P. Morgan’s Perspective sits closer to front-end analytics. Perspective’s documentation describes it as an interactive analytics and visualization component for large and streaming datasets, and a FINOS case study says it has been in production use in “1,000’s of applications” within J.P. Morgan, including trader desktops, since being open-sourced in November 2018. ### What do HRT and Two Sigma add to the picture? (developer.gs.com) Hudson River Trading’s corral is described on GitHub as “lightweight structured concurrency for C++20.” That places it in the category of reusable systems infrastructure, the kind of internal utility that can outlive any single trading strategy. Two Sigma’s flint is a Spark library for time-series analysis. Its GitHub README says it was built to analyze time-series data at scale and uses a TimeSeriesRDD structure to exploit temporal ordering and locality-based optimizations. (perspective-dev.github.io) ### What can outsiders actually learn from these repositories? The public repositories do not disclose firms’ proprietary models, positions or trading logic. They do show naming conventions, language choices, performance assumptions, interface design and the kinds of engineering problems firms decided were generic enough to open-source. (github.com) The next step is public and immediate: the repositories are available now on GitHub and, for Perspective, through FINOS documentation, where developers can review commit histories, releases and licenses project by project. (github.com 1) (github.com 2)