GroveAI
Case Study

Secure AI Data Analysis for Financial Services

Ashworth Capital manages over £4.2 billion in assets and needed AI-powered analytics without exposing client portfolio data to external services. We deployed a private AI platform that runs entirely within their secure environment.

Client: Ashworth CapitalIndustry: Financial ServicesDuration: 8 weeks

Results

Impact delivered

6x faster

Report generation speed

Portfolio risk reports generated in minutes instead of days

92%

Anomaly detection accuracy

Flagging trading pattern irregularities before manual review

Zero

Compliance incidents

No data handling violations since deployment

+58%

Analyst productivity

Measured by research output per analyst per quarter

The Challenge

What they faced

Ashworth Capital, a London-based asset management firm, wanted to use AI to accelerate portfolio risk analysis, generate investment research summaries, and detect anomalies in trading patterns. However, FCA regulations and client contractual obligations meant that portfolio data, trading records, and research notes could not leave their secured infrastructure under any circumstances. Previous attempts to use cloud-based AI tools had been blocked by their compliance team. The firm's analysts were spending days producing reports that competitors with fewer regulatory constraints were generating in hours, putting Ashworth at a competitive disadvantage.

Our Solution

How we solved it

We deployed a fully private AI analytics platform within Ashworth's existing data centre. The system uses a locally hosted Mixtral 8x22B model for natural language analysis and report generation, combined with custom-trained models for anomaly detection in trading data. Analysts interact with the platform through a secure internal web application that connects to the firm's proprietary data warehouse. The entire stack — from model inference to the user interface — runs within Ashworth's network perimeter, with no outbound data connections.

Approach

Step by step

01

Regulatory and compliance review

Worked with Ashworth's compliance team and external counsel to define the exact data handling constraints and document the system's architecture for FCA audit readiness.

02

Infrastructure provisioning

Specified and oversaw the installation of a dedicated GPU cluster (4x NVIDIA H100) within the existing data centre, optimised for large model inference.

03

Model deployment and optimisation

Deployed Mixtral 8x22B with quantisation optimisations to maximise throughput, alongside a custom anomaly detection model trained on historical trading data.

04

Data pipeline integration

Built secure ETL pipelines connecting the AI platform to Bloomberg terminal feeds, the internal data warehouse, and the portfolio management system.

05

Analyst interface development

Created an internal web application allowing analysts to query portfolios in natural language, generate risk reports, and flag anomalies — all within the secure perimeter.

06

Security hardening and penetration testing

Conducted a full security audit and penetration test with a third-party firm to validate that no data exfiltration paths existed.

We were told we'd have to choose between AI capabilities and data security. Grove showed us that was a false choice. Our analysts now have tools that rival anything the cloud-first firms are using, and our compliance team sleeps soundly knowing every byte stays in-house.

Daniel Ashworth

Chief Investment Officer, Ashworth Capital

Technology

Stack used

Mixtral 8x22BvLLMNVIDIA H100PythonFastAPIReactPostgreSQLApache Kafka

FAQ

Frequently asked questions

Market data feeds from Bloomberg and internal sources are ingested through secure, one-way ETL pipelines. The AI models process this data locally — no market data or analysis results are ever transmitted externally.

Yes. The architecture is designed to be modular. Ashworth has already begun extending access to their ESG research team, and plans to onboard the client reporting function later this year.

We provide quarterly model performance reviews and re-training cycles. Model updates are delivered via encrypted offline transfer and deployed during scheduled maintenance windows.

Want similar results?

Book a free strategy call and we'll assess how AI can deliver impact for your business.