Puurspane Real Estate Platform
Microservices platform for Puurspane uniting property management, lead CRM and geo-searchable listings with virtual tours.

Puurspane Real Estate Platform overview
Puurspane is a Real Estate application serving agents, landlords, tenants and property seekers across web and mobile. We re-architected it from a single monolithic system into a high-availability, multi-tenant cloud platform that brings property management, customer relationship tools and interactive listing operations into one decoupled ecosystem. The platform had to handle high-definition media streaming for virtual property tours, real-time notification distribution and geo-spatial listing searches on dynamic maps, all without the locking and synchronisation problems of the old system.
Our roadmap set four milestones: move all core services into a containerised microservices layout for runtime isolation; deploy an event-driven data architecture that separates read-heavy searches from write-heavy lease and financial transactions; put a centralised identity platform in front of administrative and client data; and build a continuous integration and deployment pipeline that tests and ships updates automatically, so code changes, scaling and recovery no longer require manual intervention or downtime.
Challenge in Real Estate operations
The legacy Puurspane foundation was one massive monolith in which the customer database, listing directory and property management scripts shared the same hardware. A single resource-heavy property search could slow or stop background financial tasks and lease updates through database locking. Synchronisation between business units ran on manually deployed scripts, so property availability regularly disagreed between the web and mobile views and updates arrived late.
There was no automated scaling: when listings spiked, the single database server exhausted its connection pool and both agents and clients hit request timeouts. Property photos and virtual tours lived on local server disks, capping storage and slowing media loads. With no gateway layer, every backend server was directly exposed to the public internet, which widened the attack surface and made centralised traffic monitoring impossible. Deployments were manual, technical debt was high and database queries were slow.
Solution architecture & delivery
We rebuilt Puurspane as a microservices platform on managed Kubernetes clusters spanning AWS and Azure, with Docker multi-stage images on minimal base layers and modular Terraform blueprints, using remote state locking, to define networks, security boundaries and compute. An API gateway is the single entry point for web and mobile traffic: it verifies the caller, then routes by path to the listings, property management, CRM or financial services, which run in private subnets and authenticate to one another with internal tokens. Services exchange events through RabbitMQ rather than sharing a database, and Redis holds active sessions, frequent search parameters and transient API metadata to keep load off PostgreSQL.
The property management service tracks each rental unit from initial listing to lease termination. Database triggers update status logs the moment a lease is signed; a tenant repair request is serialised to a queue, which alerts local maintenance teams and updates the landlord dashboard simultaneously without blocking read connections. The CRM service records preferences, search histories and agent interactions, and background streams watch behaviour such as saved searches or repeated views of a property style; when a user crosses an interaction threshold, the lead is passed automatically to the assigned agent and outbound follow-ups go through the message broker. Notifications for open houses or price reductions are published to decoupled queues so worker nodes can fan out push alerts and emails gradually.
The listings engine runs on PostgreSQL with PostGIS for coordinate queries, paired with Elasticsearch geo-distance and multi-filter indexes over property characteristics and text descriptions; spatial bounding-box indexes return thousands of listings within a radius in under fifty milliseconds. Images and 3D virtual tour files are stored in AWS S3 buckets, transcoded by an automated pipeline into compressed, web-friendly formats and cached on Azure CDN edges, with tiled fragments streamed according to the user's view window.
Security and operations are built into the fabric. Okta with OpenID Connect issues short-lived JSON Web Tokens validated at the gateway; administrative routes require multi-factor authentication and role-based access control. Data is encrypted in transit and at rest, sensitive documents such as income verifications and lease agreements go to a dedicated bucket with per-object keys and automated key rotation, CrowdStrike Falcon runs as a DaemonSet on every worker node, and GDPR and SOC 2 controls include non-rewritable audit logs and automated anonymisation or erasure on account deletion. The listings service scales out when average CPU exceeds sixty percent for three minutes, a synchronous PostgreSQL replica in a second zone is promoted automatically in under ten seconds on heartbeat failure, and GitHub Actions runs unit tests, syntax checks and image vulnerability scans before blue-green rollouts that monitor error rates and latency for ten minutes and revert automatically.
Key features of the Puurspane Real Estate Platform
- Property management service covering lease contracts, maintenance logs and financial transaction histories with real-time status triggers
- Queue-driven maintenance workflow: tenant repair requests alert maintenance teams and landlord dashboards simultaneously
- CRM with behavioural lead scoring: saved searches and repeat views trigger automatic hand-off to the assigned agent
- Geo-spatial listings search on PostgreSQL with PostGIS and Elasticsearch, returning radius results in under fifty milliseconds
- Virtual tour delivery: automated transcoding to S3 with Azure CDN edge caching and view-window tiled streaming
- RabbitMQ notification fan-out for open houses and price reductions without loading the core databases
- Okta OpenID Connect sign-on with MFA, short-lived JSON Web Tokens and role-based admin access
- GitHub Actions CI/CD with vulnerability scans, blue-green rollouts and a ten-minute automatic rollback window
Who this web application is for
The architecture fits property portals, letting agencies and property management companies whose listing, CRM and leasing functions have grown together into one database that locks under load. It is especially relevant where the product serves both web and mobile audiences that must see identical availability, where heavy media such as 3D tours is part of every listing, and where lease agreements and income documents impose GDPR or SOC 2 obligations on storage, access and deletion.
Impact & results
- Decoupled a monolith in which one heavy property search could stall lease updates and financial tasks into isolated, independently scaling services
- Property availability now matches across web and mobile, replacing manually deployed synchronisation scripts
- Map searches filter thousands of active listings within a radius in under fifty milliseconds
- Media moved from local disks to S3 and CDN edges, so tours load quickly regardless of distance from the data centre
- Backend servers are no longer exposed to the internet; all traffic passes an API gateway with centralised monitoring
- Failover to the secondary database completes in under ten seconds, and failed deployments revert automatically
FAQ about the Puurspane Real Estate Platform
What problem does the Puurspane Real Estate Platform solve?
It replaces a monolithic Real Estate system, in which heavy property searches locked lease and financial updates and web and mobile availability drifted apart, with isolated microservices for property management, CRM and listings that scale independently and stay consistent through event queues.
Which technologies power the Puurspane Real Estate Platform?
Kubernetes and Docker on AWS and Azure provisioned with Terraform; PostgreSQL with PostGIS for lease, financial and geographic data; Elasticsearch for listing search; Redis for caching; RabbitMQ for events; Okta for identity; CrowdStrike Falcon for runtime security; and GitHub Actions for the deployment pipeline.
How fast is the property map search?
PostgreSQL with PostGIS is paired with Elasticsearch geo-distance queries, and spatial bounding-box indexes on property coordinates let the system filter thousands of active listings within a chosen radius in under fifty milliseconds.
What happens when traffic spikes or a database fails?
When average CPU across the listings service exceeds sixty percent for three minutes, Kubernetes provisions additional containers. A synchronous PostgreSQL replica in a second availability zone is promoted automatically on heartbeat failure, rerouting connections in under ten seconds without data loss.
Puurspane Real Estate Platform product screens
Build your next Real Estate product with Next Olive
Share your requirements β we will propose scope, timeline and stack within one business day.