Track record

Eighteen years, one standard: platforms that are still running. A selection of engagements delivered by our consultants and founder — each described exactly as it happened.

REF·01
Vodafone · Telecommunications

Neuron: a group data platform, moved to the cloud without dropping a feed.

Vodafone’s group analytics estate ran on 600+ on-premises Cloudera servers across 11 countries — capacity-bound, cost-heavy, and holding back the ML roadmap. As core architects on the Neuron programme, we designed the bespoke Scala/Beam ETL framework that carried the migration to Google Cloud: 500+ daily data feeds and 88 production ML models, market by market, with the Italy migration (500+ feeds) led end-to-end. The platform also gained what the old estate never had: self-service ML execution, lineage, and proactive monitoring.

600+ SERVERS · 11 COUNTRIES
REF·02
Vodafone · Telecommunications

Margin assurance: finding the revenue leaks in billing data.

Telecom margins leak in the gaps between contracts, billing systems and network usage — invisibly, until someone builds the analytics to see them. We designed and built the margin-assurance platform (Dataflow, Dataproc, BigQuery ML) whose leakage-detection algorithms surfaced discrepancies worth millions of euros in recoveries, and added self-service ingestion so business users could bring new data sources without an engineering queue.

€M+ RECOVERED IN LEAKAGE
REF·03
Vodafone · Telecommunications

Governed data for enterprise GenAI.

An AI platform is only as trustworthy as its intake. For Vodafone’s enterprise GenAI programme we built the secure document pipeline: PGP-encrypted intake, DLP-based PII detection and redaction, HashiCorp Vault key management — with private keys and plaintext PII never persisted and never entering the processing graph. The output: governed, auditable data published for AI use, with the security team able to verify every step. The same engagement contributed to the platform’s vector stores, knowledge bases and vendor-agnostic LLM orchestration layer.

0 PLAINTEXT RECORDS EXPOSED
REF·04
Sainsbury's · Retail

Replacing a Teradata warehouse with a platform the whole group could use.

Sainsbury’s needed to move beyond a warehouse at its limits. We architected the multi-tenant enterprise data lake that replaced it — 50+ source systems and 400+ entities across retail, banking and loyalty, orchestrating 500+ daily jobs — and built the configurable Scala/Spark ETL engine and real-time ingestion framework (Kafka, RabbitMQ) that made the hundredth feed a configuration exercise rather than a project.

50+ SOURCE SYSTEMS · 500+ DAILY JOBS
REF·05
Virgin Money · Financial services

A data platform for a digital bank, built serverless from day one.

A digital bank generates data faster than a traditional data team can provision for it. We architected Virgin Money’s digital bank data lake on serverless AWS — streaming and batch ingestion from 10+ operational systems (Lambda, Kinesis, Glue, Redshift, EMR/Spark, Terraform) — feeding analytics, data science and AI workloads without a server to manage or a capacity plan to miss.

10+ OPERATIONAL SYSTEMS · SERVERLESS
REF·06
European Medicines Agency · Regulated public sector

Regulatory data platforms where an error has consequences.

Seven years designing the systems Europe’s medicines regulation runs on: adverse drug reaction processing (EudraVigilance), the agency’s data lake with signal-detection analytics, and ISO/IDMP product-lifecycle systems — serving regulators across 28 countries on Cloudera, Spark, Kafka and Elasticsearch. Regulatory-grade data handling isn’t a feature we added later; it’s where ARC’s standards were formed.

28 EU COUNTRIES SERVED
REF·07
BT · Telecommunications

Predicting broadband outages before customers feel them.

Reactive network support finds problems after the complaint. We built GCP data products analysing broadband hub telemetry to predict outages ahead of failure — Dataflow/Beam batch and streaming pipelines serving analytics for 2M+ customers across TV, broadband and voice — migrating the workload from Cloudera Hadoop with Terraform-managed infrastructure and TDD/BDD engineering practice throughout.

2M+ CUSTOMERS SERVED
END OF LINE → PRODUCTION

Want the platform behind one of these numbers?

Talk to us