Cabinet Office — Generative AI document understanding
Problem: Need to query and summarise large communications datasets safely in a regulated setting.
Action: Built an async FastAPI backend integrating OpenAI models with RAG (LangChain + ChromaDB + Redis),
added prompt-injection controls and structured testing.
Outcome: Faster discovery and summarisation workflows with stronger safety controls and a production-ready delivery approach.
RAGFastAPISecurity-aware AI
MHCLG — National housing data platform performance
Problem: ETL pipeline runtime exceeded operational tolerance (10+ hours), limiting iteration and insight.
Action: Profiled the pipeline end-to-end, introduced parallelisation and restructured bottlenecks;
demonstrated improved concurrency by migrating from SQLite to PostgreSQL.
Outcome: Reduced run time from 10+ hours to under 5 hours with clear, non-disruptive changes.
ETLPythonPostgreSQL
Met Office — HPC workflow automation & model evaluation
Problem: Scientific workflows required reproducibility, automation, and faster iteration across HPC environments.
Action: Built Python tooling for automated job submission/monitoring (schedulers, async APIs),
and improved workflow orchestration for model evaluation; introduced containerised execution for portability.
Outcome: More reproducible runs, faster feedback loops, and improved operational reliability for researchers.
HPCAutomationScientific workflows
Department for Business & Trade — System migration & integration
Problem: Required migration and integration between ticketing platforms with minimal disruption and strong auditability.
Action: Built Django REST APIs, implemented robust error handling, authentication and logging,
containerised services, and supported CI/CD with automated testing.
Outcome: Successful migration with an engineering approach aligned to regulated delivery expectations.
DjangoIntegrationCI/CD
HSBC — Model execution platform & migration support
Problem: Large portfolio of legacy models required migration and operationalisation within a new execution platform.
Action: Supported platform microservices and developed a migration framework;
worked with modellers to onboard large numbers of forecasting models and provided structured support processes.
Outcome: Improved throughput of model onboarding and strengthened operational support for a critical platform.
Platform engineeringMicroservicesRegulated finance
Deutsche Bank — Cryptography libraries for secure operations
Problem: Multiple operational and engineering teams required a consistent, policy-aligned approach to secure handling
of passwords and sensitive strings across heterogeneous systems.
Action: Designed and delivered encryption/decryption libraries as a Java JAR module and a native C shared library,
with language bindings (Perl XS module and JNI integration) to support different runtime environments;
aligned implementation with security constraints and practical operational usage.
Outcome: A reusable, multi-language cryptography capability that standardised secure handling patterns and reduced
bespoke re-implementation risk across teams.
CryptographyControlsMulti-language
Morgan Stanley — Microservices for database operations at scale
Problem: Operational burden managing large estates of legacy databases across environments.
Action: Architected and built RESTful microservices frameworks to enable self-service provisioning and maintenance;
introduced monitoring and remediation automation with audit-friendly workflows.
Outcome: Reduced manual operational load and improved reliability for a large database estate.
MicroservicesAutomationScale
FLOWGRID — EU-funded distributed computing for CFD (consortium)
Problem: CFD applications required scalable execution across distributed compute resources.
Action: Initiated and led a consortium proposal and delivery plan for a grid-based architecture;
provided technical architecture and project management across partners and academic collaborators.
Outcome: Consortium awarded €1.1m in funding; delivery progressed under programme governance suitable for multi-partner execution.
EU programmeDistributed computingCFD
Startups — Platform engineering under delivery constraints
Problem: Early-stage teams needed reliable systems without over-engineering and with rapid iteration.
Action: Delivered practical architecture, implementation, and operational hardening for data-driven services,
balancing speed with maintainability and clear technical decision-making.
Outcome: Stronger platform reliability and clearer delivery path from prototype to operational use.
StartupsArchitecturePragmatic delivery
