Goldman Sachs explains nine engineering tenets guiding data work

Goldman Sachs wants its engineers to think less like ticket closers and more like stewards of the firm’s operating system. The nine engineering tenets are not a branding exercise, they are the rules of the road for how data work gets done, who gets to make tradeoffs, and what kinds of engineering decisions earn credibility inside the firm.

What the nine tenets are really saying

Goldman publicly set out nine Engineering Tenets in a September 23, 2021 post, and says they are meant to complement the firm’s Purpose and Values while shaping how the engineering community operates. The themes are broad on purpose, but they are not vague in practice: build for clients, modernize the stack, design resilient systems, use measurable data, and foster inclusion and mutual respect.

Three tenets matter especially for data engineering: looking around corners, innovating incrementally, and inspiring trust. Taken together, they reward engineers who anticipate downstream consequences, make careful progress instead of risky rewrites, and leave behind systems that other teams can actually rely on. In a place where the wrong data decision can ripple across trading, reporting, and controls, that is not abstract culture talk. It is the difference between being seen as a cost center and being treated as part of the firm’s strategic core.

How that changes day-to-day engineering work

For product data engineers, the job is not just about moving data from one place to another. Goldman says this work sits on critical reference data systems, supports a wide range of functional requirements, and has to keep compliance obligations front and center. That means engineers are expected to think about entitlements, auditability, access restrictions, and the consequences of every schema change, not just whether the code passes tests.

The operational reality is messy. Teams are dealing with technical debt, database upgrades, unsupported systems, software library migrations, and the constant need to secure data while modernizing platforms. The tenets push engineers toward disciplined tradeoffs: modernize without breaking access, improve speed without weakening controls, and solve the immediate problem in a way that does not create a larger one for the next team.

A useful way to read the tenets is as a decision-rights guide:

• If a change improves speed but weakens trust, it is not a win.
• If a platform rewrite looks elegant but exposes clients or controls to unnecessary risk, incremental progress is the better answer.
• If a team cannot explain the measurable impact of its work, it has not yet earned the strongest version of Goldman’s trust.

That framing also changes how partner teams should work with engineering. Product owners, operations leaders, legal, and risk teams cannot treat data work as a back-end utility that simply absorbs requirements after the fact. The tenets imply that engineers should be in the room early enough to shape the design, especially when compliance, governance, and resiliency are on the line.

Legend is the clearest proof of the philosophy

Goldman’s own data-engineering materials show how long this mindset has been building. In 2021, the firm said its data-engineering team had spent over seven years building Legend, an open-source solution designed to provide accurate, timely, and safe access to data with greater efficiency and reliability. The platform was built to break down data silos, support self-service retrieval, and preserve entitlements tied to source data.

That matters because it shows the bank is not trying to win on flash. It is trying to win on infrastructure that other businesses can actually use. Goldman later said in 2024 that Legend is used for financial products, standardized reference data, market data, employee data, and more, which makes clear that this is not a niche internal tool. It is part of the plumbing that lets the firm organize and share information across the enterprise.

The open-source angle is part of the story too. Goldman said it open-sourced Legend in 2020, and the move helped turn an internal platform into something with broader industry relevance. For engineers, that signals that internal architecture at Goldman can become a product-quality asset, and for the broader firm it reinforces a simple lesson: durable infrastructure is treated as strategic intellectual capital, not invisible maintenance.

Why bankers, operations, and legal should care

If you work on the business side, the tenets tell you what kind of engineering behavior Goldman rewards and what kind it resists. Client reporting, execution, pricing, and risk all depend on stable data systems, so a “quick fix” that creates governance gaps is not really quick at all. The firm’s data platform efforts are meant to reduce those friction points, not just speed up internal work.

Goldman’s Financial Cloud for Data, launched in late 2021, makes the same point in client-facing form. The platform was described as combining modular AWS components, a managed time-series database, open-source Legend components, GS Quant, and PlotTool Pro. The Stack reported that the aim is to reduce the need for clients to build foundational data-integration technology themselves, lower the barrier to advanced quantitative analytics, and homogenize messy data from APIs, FTP, market feeds, and third parties.

That has direct implications for partner teams. Operations leaders should expect more standardized data pathways. Legal and compliance teams should expect stronger built-in governance rather than after-the-fact review. Client-facing teams should expect faster access to usable data, but only within a structure that keeps controls intact.

Governance is the feature, not the footnote

Goldman and its partners have been unusually explicit that the hard problem is not ingesting data, but governing it once it arrives. In a Snowflake partner writeup, Goldman said researchers previously had to work with data engineers to combine internal and third-party data while maintaining governance, which created bottlenecks and delays. The Legend and Snowflake workflow was presented as a way to make that collaboration self-service while keeping governance built in.

Goldman’s 2024 Google Cloud material added another layer: Legend helps unify authentication, authorization, and data-model governance so customers can interact with data more safely and efficiently. Snowflake said Goldman uses Snowflake Secure Data Sharing to onboard third-party data from vendors, sellers, and marketplaces, which underlines how much of the challenge is really about controlled access, not just storage.

For employees, that means governance is not a veto thrown on by another department. It is part of the design brief. The engineers who do best in this environment are the ones who can make controls feel native to the workflow, not bolted on after the fact.

The culture signal is bigger than one data platform

Goldman’s tenets also carry a broader culture message: mistakes are inevitable, but the expectation is to learn from them, solve issues together, and keep moving forward. The firm explicitly says it embraces failure as an essential element of growth and uses data to learn from mistakes. That is a notable stance in a business where the downside of failure can be expensive and public, and where too much caution can be its own competitive risk.

Marco Argenti, Goldman’s co-CIO, has argued that engineers need a seat at the company’s strategic table because software now shapes client experience and business agility. That is the real takeaway from the tenets and the data platform work around them. At Goldman, engineering is not supposed to sit behind the business and wait for instructions. It is supposed to help define what the business can safely do next.

The message for engineers is clear: the work is rewarded when it improves trust, keeps the firm moving, and makes complicated data usable without making it less secure. The message for everyone else is just as clear: if Goldman’s data machinery is built well, the rest of the firm gets faster, safer, and harder to outmaneuver.