What tech stack does Goldman Sachs use?

Goldman Sachs's backend is built predominantly on Java for enterprise application development — used across trading systems, risk engines, clearing platforms, and internal middleware. Python is increasingly dominant for quantitative research, data science, and machine learning applications across both GBM and AWM. C++ remains present on the highest-performance trading and database teams where microsecond latency matters, particularly in electronic market-making and high-frequency risk management. PURE, an immutable functional language developed internally and now open-sourced as part of Legend, is Goldman's canonical data modeling language, enabling consistent data definitions compilable into SQL, Java, JSON, and other downstream formats.

For cloud infrastructure, Goldman Sachs co-designed the GS Financial Cloud for Data on modular, serverless AWS components — a purpose-built architecture enabling compliant financial data sharing with institutional counterparties, meeting regulatory requirements for data residency and auditability. The firm maintains a hybrid model: cloud-first for analytics, development, and collaboration workloads, while regulated trading data and latency-sensitive systems remain on Goldman's own data centers. GitLab underpins the software development lifecycle for the Legend platform and is used broadly across Goldman's engineering organization.

Goldman's engineering headcount in Bengaluru and Dallas has grown substantially, with both locations serving as hubs for platform engineering, data infrastructure development, and internal tooling. The firm's technology investment runs approximately $4–5 billion annually across all of these domains, making Goldman one of the largest technology spenders among financial institutions globally.

The Marquee platform — Goldman's institutional digital client interface — is built on a React-based web frontend, delivering cross-asset analytics, execution tools, derivatives pricing, and market data dashboards to more than 70,000 monthly active institutional users. Marquee competes directly with Bloomberg terminal functionality for derivatives and alternatives analytics, and its API layer allows sophisticated clients to embed Goldman data and analytics into their own workflows. In 2026, Marquee Structured Products was named Best Single Issuance Platform at the SRP Europe Awards.

On data infrastructure, Goldman open-sourced Legend through the FINOS foundation in October 2020 (originally developed internally as Alloy/PURE). Legend comprises five modules — Studio, Engine, SDLC, Shared, and the PURE language itself — and has been adopted by Deutsche Bank, Morgan Stanley, RBC Capital Markets, and others for interbank collaborative data modeling, particularly around the Common Domain Model (CDM). Databricks handles large-scale data engineering pipelines and serves as the integration layer for analytics workloads; MongoDB and Elasticsearch cover flexible NoSQL query and search needs; Apache Kafka provides event-streaming infrastructure across trading and risk systems.

For GTM tooling, Salesforce is Goldman's institutional client CRM of record, managing coverage relationships across its GBM and AWM client franchises. Bloomberg Terminal and Refinitiv (LSEG Workspace) provide the critical market data feeds that underpin trading, risk, and research workflows firmwide. Workday handles human capital management and HR operations globally.

Goldman Sachs's technology posture is build-heavy and open-source-contributor-level sophisticated. The firm does not purchase commodity software when it can build proprietary infrastructure with a defensible competitive edge. This means the highest-probability vendor win categories are: specialized market and alternative data feeds that Goldman cannot generate internally; enterprise AI and ML infrastructure with a financial-services-grade compliance posture; cloud-adjacent security, observability, and cost-management tooling; risk management or regulatory reporting platforms that plug into Goldman's existing Java and Legend ecosystem; and defined-outcome ETF data and analytics (newly relevant following the Innovator acquisition).

Displacement plays are difficult. Goldman's Java, Python, and Legend stack is deeply entrenched, internally maintained, and actively contributed to the open-source ecosystem — signaling institutional commitment, not legacy debt. The most successful vendors position as additive: a new data type Goldman cannot produce internally, a new regulatory capability required by evolving rules, a new distribution channel for AWM, or a new compliance workflow. Integration stories must lead with native API support for AWS and — for maximum credibility — FINOS/Legend compatibility or the ability to ingest Legend data models.

Vendors that build Marquee integrations gain secondary distribution to Goldman's 70,000+ institutional client user base as a potential bonus benefit. Procurement cycles are rigorous: expect the VRM process to include technical security review, data residency assessment, and business continuity verification across all major office locations. Vendors who invest in a pre-sales relationship with Goldman's technology teams in Bengaluru (where pilots often originate) and New York (where decisions are finalized) achieve materially shorter sales cycles.