Danielle Fiene

Sole engineer on a pension fund modelling platform for institutional clients. Responsible for the entire stack — cloud infrastructure, backend, frontend and security architecture.

Built the full Azure infrastructure via modular Terraform: App Services, ACR, Key Vault, SQL Server, Blob Storage, RBAC, Application Insights and Log Analytics. Configured EasyAuth (Entra ID SSO) across three App Services and implemented Managed Identity token injection via pyodbc — no passwords or connection strings anywhere in the codebase. CI/CD via GitHub Actions with OIDC and Workload Identity Federation — zero static credentials throughout.

The platform consists of two separate applications: a Django/DRF runtime API serving a React demo frontend, and a standalone admin application (Vite/React + separate Django backend) through which clients can update actuarial model inputs and publish new calculation versions — without touching code. The admin workflow includes pessimistic locking, diff calculation, publish flow with Parquet-based artefact storage in Azure Blob, and automatic rollback on health check failure, with an append-only audit trail throughout.

Financial sector, production environment

Sole engineer on a multi-tenant SaaS platform built on Google Cloud Platform — cloud infrastructure, security architecture, authentication and application layer all designed and built independently.

The platform handles WhatsApp review collection for hospitality clients.

Each tenant runs in full database-level isolation via composite foreign keys; no cross-tenant data access is possible at application or database level. Onboarding uses a ticket-based signup wizard: email verification, package selection, Stripe payment setup and password creation as a sequential flow before an account is activated.

Core feature: a configurable LLM pipeline where restaurant owners select the AI model per conversation phase from a set of 12 models, with configurable token limits, cost caps, speed settings and per-phase usage metrics — all managed from a settings dashboard without touching code.

WhatsApp messages are processed via an idempotent webhook pipeline with conversation locking and exponential backoff retry.

QR codes are HMAC-signed to prevent tampering.

Data retention is GDPR-compliant with automated anonymisation after 180 days via Cloud Scheduler jobs.

Authentication runs on a dedicated SuperTokens Cloud Run service with a separate auth database.

Orphan prevention across SuperTokens and Prisma uses PostgreSQL advisory locks with compensating actions on failure.

Built geospatial map interfaces for an enterprise traffic safety and road mobility platform. Worked across the full frontend map stack — interactive layers, spatial analysis and performance optimisation.

Key work:

Mapbox vector tile server caching (24-hour TTL), base layer switcher, Nominatim and Photon geocoding with intelligent fallback across 18+ European languages, and Turf.js spatial analysis with TomTom data integration.

Reduced Google Street View API calls by ~90% via lazy loading and interaction-driven request logic. Implemented multi-tier client-side caching with TTL across multiple data sources.

Refactored core map components: throttled hover events, eliminated redundant re-renders, introduced functional state updates for stable countdowns, and consolidated duplicated translation and formatting logic into reusable functions. Reduced unnecessary API calls on initial load by ~100 requests via deferred dropdown initialisation.