Guide

SAP Databricks architect freelance: also hired as a SAP Databricks consultant freelance

As of 2026-10-05

Edited by Cédric Mary, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

A freelance SAP Databricks architect is hired to settle one question before anything is built: where SAP business data and the lakehouse meet, who governs the shared objects and who owns the business definitions. The work sits between Datasphere or Business Data Cloud on one side and a Databricks workspace on the other. It is a thin market in advertised freelance roles: on 4 October 2026 the Analytics Legends contract corpus held 103 live postings naming Databricks, 94 of them permanent, 7 freelance and 2 contract.

What the architect decides, and what a consultant builds

The title is used for two different jobs. A SAP Databricks consultant configures: a share, a catalogue entry, a notebook that reads a shared table. A SAP Databricks architect decides the boundaries those configurations sit inside, and most of the decisions are about ownership rather than technology.

Which tables stay in SAP and are shared outward, and which are produced in Databricks and registered back? Where does a business measure such as net revenue get defined, once, so that a Databricks notebook and an SAP Analytics Cloud story cannot disagree? Who operates the Databricks side when the SAP team has no data engineers? And which workloads belong on a lakehouse at all, as opposed to inside Datasphere?

The Business Data Cloud architect role has its own page on this site; this one covers the narrower seam with Databricks.

The pattern the work revolves around

Business Data Cloud is built in part on a partnership with Databricks, and its central technical promise is zero-copy sharing in both directions. SAP application data exposed as governed data products can be read from Databricks through Delta Sharing, the open protocol for sharing Delta Lake tables without moving them. In the other direction, a Delta table produced in Databricks can be registered as an external data product and consumed from SAP Analytics Cloud through a live connection, with the data staying in Databricks-managed storage.

BDC Connect for Databricks is the mechanism that links the two catalogues, and Unity Catalog is the Databricks side of the contract: a share is a named bundle of tables or views, and the provider decides per recipient which columns and rows are visible. Masking rules defined on the SAP side are meant to be enforced when the data is read from Databricks.

The architect's judgement is in when not to use it. Batch replication remains the right choice when the consumer needs data transformed or aggregated before use, or when one side cannot speak the sharing protocol. The Databricks-in-BDC pattern is the wrong one when the engagement is SAP-anchored business-user modelling and nobody will operate the Databricks side.

Who owns the semantic layer

This is the question that decides whether the architecture holds after the first year. Datasphere carries the business semantics: associations, hierarchies, currency conversion, access rules and the Analytic Models that SAC consumes. A lakehouse carries volume, history and data science. If a team rebuilds the same measures in notebooks, there are two definitions of revenue and an argument about which one the board saw.

The defensible default is that the definition lives where most consumers read it, and that anything shared to Databricks is shared as a governed data product with a named owner. The exception is a Databricks-first estate, where the task is to bring SAP data in with its meaning intact.

Decision table: when an independent fits

Use this as a screening grid for the first conversation.

Settled design, one seam missing — Fit: independent architect for a few weeks · What to ask for: A written boundary decision and an interface list · Risk: Scope grows into delivery.

Multi-source estate with Snowflake or Fabric alongside — Fit: independent, with a delivery team behind · What to ask for: A cross-platform sharing map · Risk: No one owns the build afterwards.

Databricks programme already under way, SAP data requested late — Fit: independent for a short review · What to ask for: Findings on what is shared and what is copied · Risk: Findings arrive after decisions are made.

BW estate being converted and Databricks in scope — Fit: firm or team · What to ask for: Migration waves and a staffing plan · Risk: An individual cannot carry both streams.

Pure data science on SAP extracts, no semantic need — Fit: a Databricks engineer rather than an SAP architect · What to ask for: Pipelines and notebooks · Risk: Paying an architect rate for build work.

What the market shows

Of the 103 Databricks postings, 86 also name Datasphere, 81 SAP Analytics Cloud, 63 Business Data Cloud, 54 SAP BW and 38 Snowflake. The reading is practical: in this corpus Databricks mostly appears inside an SAP analytics job, not instead of one. Germany leads with 44 postings, ahead of the United States with 12, India with 6, Switzerland with 5 and Belgium and the United Kingdom with 4 each. By seniority, 67 are senior and 22 mid-level. Twelve titles contain architect, and one of those is freelance. Source: Analytics Legends live contract corpus, public/api/contracts-lean.json, 1,610 live postings, generated 4 October 2026.

The seven freelance postings are located in Germany (3), Belgium, France, the United States and Panama. None publishes a day rate, so the corpus gives no observed rate for this work. The only published reference we hold is a modelled median of EUR 1,050 per day for cross-stack work, meaning SAP combined with Databricks, Fabric or Snowflake, at France and 6 to 10 years of seniority, gross to the consultant, as at 8 May 2026 (Analytics Legends rate calculator, an editorial band model, not an observed population).

Independent or firm

An independent suits a bounded question: a boundary decision, a review of a sharing design, a memo that survives a change of team. A firm suits a programme with several streams. Ask who will operate what is designed; a design nobody operates is the usual way these engagements fail.

What we cannot assert

No Databricks posting in the corpus publishes a day rate, so what an independent earns on this seam is not measured here. The cross-stack figure is a modelled median. Titles do not reveal how many postings called consultant are architect roles in practice, nor how many of the 103 postings sit in SAP-anchored versus Databricks-first estates.

Frequently asked

What does a freelance SAP Databricks architect do?

Decides where SAP data and the lakehouse meet: which objects are shared through Delta Sharing and BDC Connect, which are copied, who governs them in Unity Catalog and where the business definitions live. Build work is usually left to a consultant or a data engineer.

Is a SAP Databricks consultant the same as an architect?

No. The consultant configures and builds inside a design; the architect sets the boundaries. Titles are used loosely, so ask for a boundary decision they owned and a build they completed.

Are Databricks architects with SAP knowledge hired freelance?

Rarely in advertised roles: 7 of 103 live Databricks postings on 4 October 2026 are freelance, against 94 permanent. Independent work usually comes through direct or intermediary conversations that postings do not show.

What this page is built on

BDC Connect for Databricks
Databricks-in-BDC Lakehouse Pattern
Databricks Unity Catalog — Data Sharing (External + Internal)
SAP Datasphere vs Databricks — Decision Frame for the SAP Data Platform
Live contract corpus, 4 October 2026