Data platform services
Design, build and run the data platform your institution depends on: the pipelines, databases and reporting that run the business today, and the AI data layer on top of them. On premises, sovereign, and supported for the years after go-live.
One platform for the reports you file today and the AI you want tomorrow
A bank's data already flows through dozens of pipelines: the end-of-day batch, returns to Bangladesh Bank, reconciliation, risk and finance reports. A ministry or a force has the same pattern with records, payments and logistics. Every AI initiative draws on that same data, and most stall when the data turns out to be stale, duplicated or impossible to govern.
We design, build and run the whole platform as one system: streaming and change data capture from core systems, databases and caches, a lakehouse that holds one governed copy of the data, and the document pipelines and vector indexes that feed assistants and agents. It runs in your data centre on Red Hat OpenShift, on servers and storage from Dell, HPE and NetApp, with Oracle where the core demands it.
Our engineers stay after go-live: monitoring, disaster recovery drills, upgrades and support, with your own team trained to take over.
- Years in business
- 35+
- Certified engineers
- 50+
- Projects delivered
- 200+
- Managed operations cover
- 24/7
Signs you need a platform, not another extract
Most organisations grow their data estate one report and one tool at a time. These are the signs it is time to run it as one platform.
Reports disagree
Finance, risk and the regulator's return show different numbers for the same day, because each team keeps its own extract.
The AI pilot stalled on data
The assistant worked in the demo, then failed in production because the documents were stale, duplicated or not allowed to be shared.
Everything waits for the night batch
Dashboards and fraud checks run on yesterday's data, so decisions are a day late.
One vendor owns the data
Formats and licences tie the data to a single product, and every change goes through that vendor.
Nobody can trace a number
There is no catalogue or lineage, so an audit question about where a figure came from takes weeks to answer.
Recovery is untested
Backups exist, but the database has never been failed over to the second site with the applications running.
Our data platform services
From the first assessment to running the platform every day. Use one service, or all of them.
Assessment and roadmap
Find out whether your data is ready for reporting and AI, and leave with an architecture, a budget and a roadmap your board can approve.
Learn more
Architecture and design
A reference architecture sized for your data, with DC and DR, security, governance and the vendor choices to match.
Learn more
Streaming and integration
Kafka, change data capture and APIs, so events from core systems reach fraud checks, dashboards and AI in seconds.
Learn more
Databases and caching
Oracle, PostgreSQL, Redis and Data Grid on OpenShift, replicated to a second site and failed over on a schedule.
Learn more
Lakehouse and analytics
One governed copy of the data in open table formats, queried for regulatory returns, dashboards and models alike.
Learn more
AI data platform
Document pipelines, vector search and retrieval with access labels, on your own GPUs or a vendor data platform.
Learn more
Managed operations
Monitoring, disaster recovery drills, upgrades and support under agreed service levels, with your team trained to take over.
Learn more
Database modernisation
Move databases off costly licences and old versions onto the platform, with a cutover you can reverse.
Learn more
AI applications on top
Assistants, document intelligence and agents built by our AI team on the data layer, with governance built in.
Learn moreOne pipeline, two outputs
Reports and AI drawn from the same governed data
The pipelines that feed regulatory returns and dashboards are the same ones that feed models, assistants and agents. We build them once, with quality checks, lineage and access control shared by both paths, so an AI answer is as current and as correct as the report it sits beside.
- Sources
- Core systems, databases, logs and documents, read where they are.
- Streaming and batch
- Change data capture and Kafka for what must be fresh; scheduled jobs for the rest.
- Lakehouse and index
- One copy of the data in open table formats, plus the vector and keyword index for documents.
- Reports and AI
- Regulatory returns and dashboards on one side; models, assistants and agents on the other.
Platform routes
Four ways to build it
The right route depends on where the data may live, the skills you have and the support you can get. We build and run all four, and show you the comparison. Upkeep shows how much running each one asks of your own team.
01
Open source on OpenShift
Kafka, PostgreSQL, Iceberg, Trino and a vector database, run as a platform.
Open formats and no licence lock-in, with Red Hat support underneath and our engineers on the components. The data stays portable to any future platform.
Best for: Institutions that want control of the stack and the lowest licence cost.
UpkeepHigh02
Vendor data platform
Dell AI Data Platform, HPE, NetApp or VAST, with their data engines.
Storage and data engines from one vendor, under one support contract, built on the same open-source engines underneath. Faster to procure, and tied to that vendor's roadmap.
Best for: Organisations that prefer one product and one vendor to call.
UpkeepMedium03
Oracle-centred
Oracle Database, Exadata and GoldenGate around an Oracle core.
Vector search inside the database, change data capture with GoldenGate and an open lakehouse alongside, for institutions whose core systems already run on Oracle.
Best for: Banks with Oracle core banking and the licences to match.
UpkeepMedium04
Hybrid with the cloud
A sovereign core in your data centre, with cloud services for what may leave.
Customer data and models stay in the country; analytics on anonymised data or burst capacity can use the cloud, under rules you set and we enforce.
Best for: Organisations with mixed data classifications and some cloud approval.
UpkeepHigh
How a data platform programme runs
Advise, design, build and run: the same four phases we use across all our work, applied to data.
01
Advise
Assess the data, the pipelines and the rules you work under, agree the first use cases, and set a roadmap and budget.
02
Design
Draw the reference architecture, size it, design DC and DR, security and governance, and choose the vendors to match.
03
Build
Build pipelines, databases, the lakehouse and the AI data layer as code on OpenShift, test them with your data, and move them into service in waves.
04
Run
Run the platform under agreed service levels, rehearse recovery, upgrade on a schedule and train your team to take over.
How we build data platforms that last
A data platform outlives every application that uses it. These practices keep it correct, portable and affordable to run.
01
Open formats first
Data lives in Iceberg tables, Parquet files and PostgreSQL, so it can be read by any engine and moved to any platform later.
02
Governance that reaches the answer
Access rules, lineage and retention follow the data into the lakehouse and the vector index, so an assistant never quotes a document its user may not open.
03
Real time where it matters
Change data capture and streaming replace the night batch where decisions need current data; scheduled jobs stay where they are enough.
04
Sovereign by design
The platform runs in your data centre, and air-gapped where the data must not leave the building, with no dependency on an outside service.
05
Built for your team to run
We document, pair and train as we build, so your engineers can operate the platform with us or without us.
06
Measured by freshness and correctness
We agree what matters, such as data freshness, reconciliation breaks and recovery time, then measure and report it.
Technology we work with
We are not tied to one vendor. We choose each component for support, skills and fit, and keep the list short.
Streaming and integration
- Apache Kafka
- Red Hat Streams for Apache Kafka
- Debezium
- Oracle GoldenGate
- Apache NiFi
- WSO2 Micro Integrator
Databases and caching
- Oracle Database
- Oracle Exadata
- PostgreSQL with Crunchy Data
- Redis
- Red Hat Data Grid
Lakehouse and analytics
- Apache Iceberg
- Trino and Starburst
- Apache Spark
- Apache Airflow
- dbt
- Apache Superset
- Grafana
AI data
- Docling
- Milvus
- pgvector
- Elasticsearch and OpenSearch
- NVIDIA NeMo Retriever
- NVIDIA RAG Blueprint
Platform and storage
- Red Hat OpenShift
- Red Hat OpenShift AI
- NVIDIA AI Enterprise
- Dell PowerScale and ObjectScale
- HPE Alletra Storage
- NetApp ONTAP and Trident
Governance and observability
- OpenMetadata
- OpenLineage
- Apache Polaris
- Keycloak
- Prometheus
- Grafana
- OpenTelemetry
Partners we build on
Data platforms for the sectors we know
Each sector has its own rules on where data may live, how long it is kept and who may see it. Our sector pages show where a data platform fits.
Banking and finance
Regulatory returns, fraud detection and customer analytics from one governed copy of core banking data, with DC and DR.
Government and defence
Sovereign, air-gapped platforms for national records, case systems and the document archives behind AI assistants.
Healthcare
Patient records, imaging and laboratory data integrated under strict access control, for reporting and clinical AI.
Education and research
Research data lakes and shared GPU platforms on open foundations, built for many groups to use.
Development partners
Programme data from the field to the donor report, with lineage the auditors can follow.
Ways to work with us
Data platform assessment
A review of your data, pipelines and rules, ending with an architecture, a budget and a roadmap.
Build project
We design and build the platform against an agreed scope, timeline and acceptance tests, then hand it over.
Managed operations
We run the platform day to day under agreed service levels, with 24/7 cover where you need it.
Engineers in your team
Data and platform engineers who join your team, pair on real work and train your people.
Common questions
We already have a data warehouse. Do we need a lakehouse?
Not always. A warehouse that serves your reports well can stay. A lakehouse earns its place when documents, logs and streams have to sit beside the tables, when AI needs the same data, or when the warehouse licence is the reason nothing new gets built. The assessment compares the two on your workloads.
Lakehouse and analyticsCan the platform run without an internet connection?
Yes. We build air-gapped platforms for institutions whose data must not leave the building: the software is mirrored inside, updates arrive through a controlled process, and nothing depends on an outside service. Our lab tests every component this way before it reaches you.
Architecture and designShould we buy a vendor's AI data platform or build on open source?
Both work, and we build both. A vendor platform gives you one product and one support contract; the open-source route gives you lower licence cost and data that moves anywhere. The vendor engines are built on the same open-source projects underneath, so the skills carry over either way. We show you the comparison for your case.
AI data platformHow does this connect to our AI plans?
Directly. Assistants and agents are only as current and as correct as the data behind them. The same pipelines that feed your reports feed the document index and the model inputs, with the same access rules. Our AI team builds the applications on top.
AI and machine learningCan we keep Oracle?
Yes. Many of our banking clients run Oracle core systems, and the platform is designed around them: change data capture with GoldenGate or Debezium, Oracle's own vector search where it fits, and an open lakehouse alongside for everything else.
Databases and cachingHow long does it take?
An assessment takes weeks. The first pipelines and the lakehouse usually go live within a few months, starting with one use case that has a measurable result. The platform then grows one feed and one use case at a time, which is also how the budget is approved.
Plan your data platform
Tell us what you run today and what you want the data to do. We'll suggest a first step, whether that is an assessment, a new platform or support for the one you have.