PRJ_008
008
Frontend Architecture · Fintech · React/TypeScript
Mantarisk Risk Assessment UI
Mantarisk AG · Zug, Switzerland · Apr, 2024 – Mar, 2026
Overview

Led the full migration from a no-code platform to a production-grade React and TypeScript frontend for a Swiss risk assessment and portfolio analytics tool. Delivered a reusable component system, real-time risk metric dashboards, and a significantly improved performance profile.

The Problem

No-code platforms offer speed early and cost flexibility late. Mantarisk had hit the ceiling — feature velocity had slowed, performance was degrading under data load, and the frontend had become a blocker for the product roadmap. The challenge wasn't just technical migration; it was building enough trust to justify rewriting something that was, technically, working.

Key Metrics
60%
Bundle size reduction
3×
Feature delivery speed
<1.2s
Dashboard load time
100%
Component reuse rate
Process & Timeline
Week 1–2
Audit & architecture planning
Audited the no-code platform capabilities and mapped every feature to a React equivalent. Identified performance bottlenecks in the existing data visualisation layer.
Month 1
Design system foundation
Built the component library ground-up — tokens, typography, spacing, colour system. Established strict TypeScript interfaces for API contracts.
Month 2–3
Core feature migration
Migrated risk assessment forms, portfolio views, and metric dashboards. Integrated REST APIs for real-time portfolio data with optimistic UI updates.
Month 4+
Performance & handoff
Webpack bundle analysis, code splitting, lazy loading. Reduced initial bundle by 60%. Documented component API surface for the wider team.
Tech Stack
ReactTypeScriptWebpackREST APIsCSS ModulesChart.js
Critical Self-Evaluation
The no-code to coded migration was fundamentally a trust problem before it was a technical one. I had to demonstrate early wins — the bundle size reduction metric was the one that won the internal argument, and I deliberately prioritised it in the first sprint for that reason. That's a lesson in knowing which metric your stakeholders care about and delivering it first. What I underestimated was the time cost of feature parity work — migrating 'what exists' before building 'what's better' is a psychological challenge as much as an engineering one. I'd negotiate a harder cutover date next time rather than running parallel systems.