ActuEdge

Most apps tell you what your money did yesterday. ActuEdge uses AI and the risk science behind billion-pound pension funds to show you what it might do tomorrow - and what to do about it.

Comment

Inspiration

I've spent over 15 years in financial services - auditing insurers at KPMG under Solvency II, implementing risk models at Moody's Analytics, and now leading actuarial validation on multi-million pound system migrations. Every day I work with tools that stress-test portfolios worth billions: Value at Risk models, Monte Carlo simulations, liability matching frameworks, stochastic reserving.

Then I open my personal investing app and get… a pie chart and a percentage.

The gap is staggering. The same risk science that protects pension funds and insurance companies is completely absent from retail investing. Everyday investors are flying blind - they can't see hidden correlations between their assets, they don't know how their portfolio would survive a 2008-style crash, and nobody is matching their investments to their actual life goals like retirement or a house deposit.

I'm also currently studying an MSc in Actuarial Management with Data Science at Heriot-Watt University, where my coursework on neural networks and machine learning gave me a second realisation: AI can make this institutional-grade intelligence accessible to everyone, not just City professionals.

ActuEdge was born from that frustration and that opportunity.

What it does

ActuEdge is a portfolio tracker that gives everyday investors the same risk intelligence that billion-pound institutions use - wrapped in a beautiful, intuitive interface.

Core Features: Dashboard - Real-time portfolio overview with performance vs benchmark, asset allocation breakdown, and KPI cards showing total value, P&L, and income yield.

Actuarial Risk Engine - Value at Risk calculations at 95%, 99%, and 99.5% confidence intervals. Portfolio health radar comparing your diversification, liquidity, volatility control, and income yield against optimal targets. Concentration and currency risk gauges.

ML Regime Detection - A neural network (LSTM + Attention architecture) classifies the current market as Bull, Bear, Sideways, or Volatile using 54 macro and technical features. The regime dynamically adjusts your risk scores and generates actionable recommendations like "increase hedge ratio" or "shorten bond duration."

Correlation Heatmap - Interactive 10×10 matrix showing how every asset in your portfolio moves relative to every other. Flags hidden concentration risk - for example, owning Apple and an S&P 500 ETF feels diversified but they're 89% correlated, effectively doubling your large-cap tech exposure.

Monte Carlo Projections - 5,000 stochastic simulations projecting your portfolio over a 5-year horizon, displayed as a fan chart with P20 through P80 confidence bands. Plus historical stress tests replaying the 2008 GFC, COVID crash, rate shocks, and stagflation scenarios.

Liability-Driven Goal Matching - Set future goals (retirement, house deposit, children's education) and ActuEdge matches your asset durations to your liability durations using actuarial immunisation principles. Shows your funding ratio, duration gap, and exactly which assets to add to close the gap.

Tax-Loss Harvesting - Automatically flags positions with unrealised losses near tax year-end, calculates your potential CGT saving at 20%, and suggests smart swap trades that crystallise the loss while maintaining market exposure. Includes HMRC Bed & Breakfasting rule compliance tracking.

Open Banking Integration - Auto-imports holdings from UK platforms via Plaid and TrueLayer. Connects current accounts, ISAs, SIPPs, and trading platforms in one view. FCA regulated, PSD2 compliant, read-only access.

Smart Alerts - Bond maturity reminders, coupon payment schedules, concentration warnings, regime shift notifications, and tax harvest windows - all prioritised by urgency.

How we built it

Frontend: React with Vite as the build tool. The entire application is a single-page app using functional components and React Hooks (useState, useEffect, useMemo). All styling is inline for maximum portability - no CSS framework dependency.

Data Visualisation: Recharts library for all charts - area charts for portfolio performance, pie charts for allocation, radar charts for portfolio health, bar charts for regime history, and a custom-built interactive correlation heatmap using pure React.

Design System: A luxury dark fintech aesthetic inspired by Bloomberg Terminal meets premium consumer brands. Custom colour palette anchored by dark backgrounds and gold accents . Animated transitions, glow orbs, and subtle hover effects throughout.

Risk Models: The actuarial calculations - VaR at multiple confidence intervals, Monte Carlo fan charts, duration matching, and stress test scenarios - are all modelled from real institutional frameworks I use in my professional work. The correlation matrix uses realistic pairwise correlations based on actual asset class behaviour.

ML Architecture: The regime detection model is designed as an LSTM with temporal attention, trained on 54 input features spanning macroeconomic indicators, technical signals, sentiment data, and flow data with a 120-day lookback window.

Challenges we ran into

Translating institutional concepts for retail users. In my day job, I can say "your duration gap implies a negative DV01 mismatch" and everyone understands. For ActuEdge, I had to make liability-driven matching feel intuitive - so instead of actuarial jargon, users see "Your retirement goal is 35% funded. Add longer-dated bonds to close the gap." Finding that balance between rigour and accessibility was the hardest design challenge.

Fitting 10 feature-rich tabs into a single React component. Keeping the codebase manageable while building a dashboard, holdings table, risk engine, ML regime tracker, correlation heatmap, Monte Carlo projections, goal matching, tax harvesting, open banking hub, and alerts system - all with consistent styling and smooth transitions - required careful component abstraction and data architecture. Making the correlation heatmap interactive and readable. A 10×10 matrix with hover states, colour-coded correlation values, and dynamic alert panels needed careful state management and performance tuning to feel smooth rather than laggy.

Realistic financial modelling in the browser. The Monte Carlo simulations, VaR calculations, and stress test scenarios needed to be computationally lightweight enough for client-side rendering while still producing actuarially meaningful outputs.

Accomplishments that we're proud of

10 fully functional tabs - each one could be a standalone feature in a competing app. Together they create something genuinely new in the retail investing space.

Bridging two worlds authentically. This isn't a developer guessing what actuarial science looks like - it's built by someone who's spent 15 years doing this work professionally at KPMG, Moody's, and Standard Life Aberdeen, now adding ML/AI skills through a postgraduate programme at Heriot-Watt.

The correlation heatmap with hidden concentration alerts. No retail investing app shows users that their "diversified" portfolio of Apple + S&P 500 ETF is actually 89% correlated. This single feature could prevent real financial harm.

Liability-driven goal matching for consumers. This concept is the backbone of how pension funds manage billions, but has never been available to individual investors. ActuEdge makes duration matching and funding ratios feel as natural as checking your bank balance.

The design. Financial tools don't have to look like spreadsheets from 2005. ActuEdge proves that institutional-grade analysis can feel premium and beautiful.

What we learned

Actuarial science and ML are natural partners. Regime detection improves every downstream calculation - VaR means something different in a Bull vs Volatile market. Connecting the ML layer to the actuarial layer creates compounding value.

Users don't want more data - they want better decisions. Early iterations showed too many numbers. The breakthrough was shifting from "here are 20 metrics" to "here's what you should do about it" - regime-adjusted actions, specific bond recommendations for duration gaps, named swap suggestions for tax harvesting.

Design is a feature, not decoration. The gold-and-dark aesthetic isn't just pretty - it builds trust. When you're asking someone to make financial decisions, the interface needs to feel as authoritative as the maths behind it.

The gap between institutional and retail tools is a massive opportunity. Everything in ActuEdge exists in institutional platforms that cost £50,000+ per year. The technology to deliver it for £4.99/month has been available for years - the missing piece was someone who understands both worlds.

What's next for ActuEdge

Near-term (3 months): Live Open Banking integration via Plaid and TrueLayer APIs for automatic portfolio syncing Real-time market data feeds for listed assets Push notifications for regime changes, tax harvest windows, and bond maturities Mobile-responsive PWA so it works on any device

Medium-term (6 months): Full ML pipeline deployment - train the LSTM regime classifier on live data and serve predictions via API AI-powered portfolio rebalancing suggestions personalised to each user's goals and risk tolerance PDF report generation for tax advisors and financial planners Multi-currency support for international portfolios

Long-term (12 months): Launch as a freemium SaaS product (Free / Premium £4.99/mo / Pro £9.99/mo) FCA regulatory sandbox application for personalised financial guidance B2B white-label offering for IFAs and wealth managers who want actuarial intelligence for their client dashboards Seek £250K pre-seed funding to build out the team and scale

The vision is simple: every investor deserves the same risk intelligence that protects the world's biggest pension funds. ActuEdge makes that real.

Built With

Share this project:

Updates