Case Studies

Selected work.

A small sample of recent engagements. Some details are anonymised at the client's request. We are happy to provide references on request.

Professional Services

Compliance & Advisory

Regulatory Horizon Scanner

Situation: A professional services firm's compliance team was spending significant time each week manually scanning news sources across UK, US, and EU jurisdictions for regulatory developments relevant to their clients.
Task: Automate multi-keyword regulatory news monitoring across jurisdictions and produce AI-generated digests suitable for direct client briefing.
Action: Built a FastAPI backend that fetches multiple News and Regulatory RSS feeds per region and keyword, then calls Claude to produce concise regulatory digests per topic. A no-install browser application runs parallel scans across all keywords simultaneously and generates PDF briefings client-side.
Result: The compliance team now runs a full multi-keyword scan across four jurisdictions in under a minute. AI-generated digests replace manual article review, with PDF export enabling briefings to be shared directly from the tool.

Financial Technology

Proprietary Trading

Situation: A systematic trader managing a portfolio of OKX inverse perpetual futures was manually reviewing funding rates and momentum data across multiple assets - a slow, inconsistent process that made timely position decisions difficult.
Task: Build a quantitative signal system that blends momentum and funding rate analysis into a single actionable position score per asset, with on-demand AI interpretation of the signals.
Action: Built a FastAPI backend computing a composite score that blends time-series momentum across four lookback periods with carry, crowding, and confirmation signals from the funding rate. A server-side Claude integration streams contextual analysis using a curated pattern catalogue. A standalone browser dashboard visualises the full portfolio and surfaces AI analysis per asset on demand.
Result: The full portfolio can be reviewed in seconds, with Claude surfacing actionable pattern insights on demand - replacing hours of manual chart and data review each week.

Financial Services Recruitment

Post-trade & BD Intelligence

Situation: A post-trade financial services firm needed to track competitor intelligence - personnel moves, M&A activity, and strategic developments - across their sector. Manually monitoring trade press and announcements resulted in patchy, delayed coverage that business development teams could not rely on.
Task: Build a market intelligence platform that aggregates post-trade sector signals - personnel moves, deals, and strategic announcements - and surfaces what matters to the Business Development (BD) team without manual monitoring effort.
Action: Built a browser-based dashboard that monitors configured target companies across a curated source register of trade press and announcements. Stories are tiered by significance with drill-down detail cards per event. A Claude synthesis pass distils raw intelligence into structured briefings, and an ad-hoc scan mode allows targeted research against specific companies on demand.
Result: The BD team receives structured, tiered intelligence on competitor moves and market signals without manual monitoring - with AI synthesis surfacing the highest-priority developments for immediate action.

Investment Management

Workflow Automation

Situation: An investment manager needed a consistent weekly briefing on scheduled macro volatility windows - central bank events and high-impact data releases across the US, Japan, Eurozone, and UK that affect rates, equities, crypto, and gold simultaneously. Producing this each Sunday required significant research across multiple sources in different time zones.
Task: Automate the Sunday digest - source aggregation, BST time conversion, cross-asset event tiering, formatted output, and email drafting - with no backend server or bespoke application code.
Action: Encoded the entire workflow as a SKILL.md - a structured, version-controlled document specifying data sources, tiering logic, time zone conversion rules, output format, and edge case handling. When invoked each Sunday, Claude reads the skill, fetches live calendar data, generates a macro backdrop from current policy rates, classifies events by cross-asset impact, renders a formatted weekly widget, and creates a Gmail draft - all in a single pass. The workflow logic lives in a file, not in application code, so it can be refined like any other document.
Result: A consistent, high-quality weekly digest is generated every Sunday in minutes. Because the intelligence lives in SKILL.md rather than locked into code, the client can update sourcing rules, tiering criteria, or output format themselves - with no developer involvement.