Lakehouse foundation and platform architecture
Workspace topology, storage design and the pipelines that keep it current — built once, documented, and handed over.
- Delta Lake
- Lakeflow
- Medallion
- Serverless
- Unity Catalog
Start your data and AI journey
with Databricks
A Bronze-tier Brickbuilder partnership, five engagements you can scope separately, and a lakehouse governed well enough to put in front of a regulator.
Talk to a Databricks expert
Bronze-tier partner · Brickbuilder Partner Network
The partnership
Build and run apps, agents and AI on your data
Marzal Labs and Databricks partner to bridge the gap between data strategy and enterprise AI. We combine the Data Intelligence Platform with delivery experience earned inside regulated industries, so the move to a unified lakehouse is one you can still run in five years. Our architects engineer resilient systems that operationalise generative AI and advanced analytics — governed, high-performance architectures rather than an implementation that stops at go-live. Together we turn critical data initiatives into a trusted production reality.
20+ yrsinside regulated data estatesLloyd’s of London, the NHS, FCA-regulated firms and UK government
21named assets already builtProducts, diagnostics, agent patterns and governance frameworks
5Databricks engagementsEach scoped, run and judged on its own rather than sold as a programme
Bronzetier in the Brickbuilder Partner NetworkWith production delivery on Fabric and Snowflake alongside it
The Data Intelligence Platform
Four layers, and they stack
Databricks unifies the warehouse and the lake into one open platform, then puts generative AI on top of it. Read it bottom up: the lakehouse foundation carries everything above it. We implement all four and hand them over documented.
- 01
Databricks SQL & Genie
Governed metric views and natural-language analytics, so the business can ask a question without raising a ticket.
- 02
Mosaic AI
Custom models, Vector Search, Model Serving and agents that ship with a measured evaluation rather than a demo.
- 03
Unity Catalog governance
One permission model across every table, file, volume, model and dashboard, with end-to-end lineage that holds up in an audit.
- 04
Lakehouse foundation
Delta Lake storage and Lakeflow pipelines — ingestion, declarative transformation and jobs. One copy of the data, reliable enough to run the business on.
Migration and modernisation
Legacy warehouses, Hadoop estates and other clouds moved onto the Data Intelligence Platform — with a reconciliation trail.
- Lakebridge
- Teradata
- Hadoop
- Snowflake
- Synapse
- Hive metastore
Unity Catalog governance, security and audit
One governance layer across data and AI assets, producing the evidence a regulator or internal audit will actually ask for.
- Unity Catalog
- Row filters
- Column masks
- Lineage
- System tables
- Delta Sharing
Mosaic AI — agents, models and evaluation
Document extraction, classification and triage agents, each shipped with an evaluation harness and documented failure modes.
- Mosaic AI
- Vector Search
- MLflow
- Model Serving
- Agent Bricks
- AI Functions
Databricks SQL, Genie and AI/BI
Governed metrics and natural-language analytics — built on a semantic layer, because a chat box over ungoverned tables only produces confident wrong answers faster.
- Databricks SQL
- Genie
- Metric Views
- AI/BI
- Lakebase
Agentic data — assistants over governed sources
Knowledge Assistants that answer from your documents, and Genie that answers from your tables — both inside Unity Catalog permissions.
- Agent Bricks
- Knowledge Assistant
- Genie
- Vector Search
- AI Functions
Agentic flow — supervised multi-agent routing
A Supervisor Agent routing each question to the right specialist, with the routing rules written down and testable.
- Supervisor Agent
- Multi-agent
- UC Functions
- MCP
- Model Serving
Why Marzal Labs
What you can expect, and what you can hold us to
- Trust, because the work carries consequence:
Every output explained, sourced and traceable, and every regulated decision reviewed by a person who is accountable for it. The governance model, the PII handling and the audit evidence are designed for firms that answer to the FCA, Lloyd’s or NHS information governance.
- Experience earned inside regulated data estates:
Twenty years across Lloyd’s of London, the NHS, FCA-regulated firms and UK government programmes. Control requirements, change governance and evidence expectations are familiar ground rather than something we adapt to on your project.
- Expertise with the artefacts to prove it:
Unity Catalog topology, row filters, column masks and lineage specified in the first architecture session. Agents built as single-purpose steps with labelled test sets and documented failure modes. Migrations reconciled row-level and aggregate before cutover, so finance can prove the numbers still match.
- A senior team, with no bench between you and them:
Four practitioners across architecture, engineering and delivery. The people who design the platform are the people who build it, and runbooks and training are part of the engagement — the platform is meant to outlast us.
- And a Bronze-tier Brickbuilder partnership, with a straight answer on fit:
We hold a Bronze-tier enrolment in the Brickbuilder Partner Network. Where Microsoft Fabric or Snowflake suits your estate better, that is what we will recommend — we have delivered production platforms on all three.
Who we are
The people who take a Databricks platform into production
Marzal Labs is a data and AI consultancy for regulated industries and a Bronze-tier partner in the Databricks Brickbuilder Partner Network. We combine lakehouse architecture, engineering and delivery judgement earned inside Lloyd’s, the NHS, financial services and government — and we hand the platform over documented, so your team can run it without us.
Meet Marzal Labs20+Years inside regulated data estates — Lloyd’s of London, the NHS, FCA-regulated firms and UK government programmes
3Products live in production on client tenancies, built by the same people who architect the platform
4Senior practitioners across architecture, engineering and delivery, with no bench between you and them
Proof from the platform
What insurers and financial firms are doing on Databricks
Published by Databricks — their customers, their figures, not ours. They are here because they are the closest public analogues to the engagements above.
All Databricks customer storiesDatabricks customer story
Allianz UK Personal Lines: a unified data platform for faster, smarter insurance decisions
Read on databricks.comDatabricks customer story
Storebrand: analytics and ML innovation at speed and scale
Read on databricks.comDatabricks customer story
Highstreet Insurance Partners: one connected view across 200+ independent agencies
Read on databricks.comTrusted clients
Trusted where the work carries real consequence
Related articles from Databricks
Straight from the people who build the platform
Published on the Databricks blog. Useful background on the governance and AI layers the engagements above are built on.
The Databricks blog