Semantic Engine — UrbanPillar & CornerSpaces

UrbanPillar and CornerSpaces are property-tech products that need listings, search, and content to actually understand what a user is looking for — not just match keywords. At Kaycomm Services, I built the semantic engine behind that: search, recommendations, and content understanding, plus an AI-driven workflow website builder that automates page and content generation, and the LLM data pipelines feeding both.

Semantic search over property listings has to reconcile two things at once: understanding unstructured intent ("3BHK near a good school") and structured constraints (price, location, layout) — while staying fast enough for an interactive product. On top of that, the workflow website builder has to generate correct, on-brand pages automatically, not just plausible-looking ones.

Clients talk to a Node.js/Express API layer. Auth runs through JWT plus Google OAuth and Firebase phone/OTP. PostgreSQL is the system of record; Redis backs caching, sessions, and BullMQ background jobs; Socket.IO handles real-time events (live listing updates, chat). Media goes to AWS S3 and Cloudinary, payments through Stripe, maps through Mapbox. The semantic engine sits behind that API layer: listing and content data gets processed through an LLM data pipeline (topics, blog content, AI-generated images) and indexed for semantic retrieval, which powers search/recommendations and a Vastu-analysis module that interprets floor plans for spatial reasoning. The workflow website builder consumes the same content pipeline to automate page generation. The whole stack — API, data pipelines, and the company's internal HRMS system — runs containerized on a VPS behind NGINX / Nginx Proxy Manager, managed through Docker + Portainer, with CI/CD automation handling deployment and domains/SSL/reverse proxy owned end-to-end.

• Built the semantic engine that powers intelligent search, recommendations, and content understanding across UrbanPillar and CornerSpaces.
• Built an AI-driven workflow website builder that automates page and content generation.
• Created LLM data engine pipelines for topics, blog content, and AI-generated images.
• Implemented a Vastu-analysis module using floor-plan interpretation and spatial reasoning.
• Authentication layered across JWT sessions, Google OAuth, and Firebase phone/OTP for web and mobile sign-in.
• Background jobs (notifications, media processing, scheduled tasks) moved off the request path into BullMQ workers backed by Redis; real-time features (chat, live listing status) run over Socket.IO instead of polling.
• Media uploads routed through S3 / Cloudinary and payments through Stripe, keeping the API stateless and PCI scope out of the app server.
• Developed core modules for the company's internal HRMS system.
• Own full-stack Docker + Portainer deployments, CI/CD automation, and performance optimization; manage VPS, domains, SSL, and reverse proxy for production infrastructure.

• Background jobs run through BullMQ with retry policies instead of fire-and-forget calls.
• NGINX / Nginx Proxy Manager terminate TLS and isolate services behind a reverse proxy.
• CI/CD automation catches build failures before they reach production containers.
• Docker isolation keeps services (API, workers, pipelines, HRMS) independently restartable without taking the whole stack down.
• Domain, SSL, and infrastructure ownership means production issues get diagnosed and fixed end-to-end, not handed off.

A live, multi-service platform in production — handling authentication, real-time messaging, media, payments, and maps — with the semantic engine, AI workflow website builder, and LLM data pipelines running on top of it for UrbanPillar and CornerSpaces.