Search Infrastructure Architect – Elasticsearch, PostGIS, & Uber H3 Expert (Consultant/Lead)

We are building a next-generation business directory platform. We are bootstrapping our database with a backlog of 5 million core business listings. We have an internal team of 2-3 capable MERN stack developers to handle frontend React components, standard API endpoints, and general Node.js routing. However, we need a specialized Search & Geolocation Infrastructure Architect to act as a consultant/lead. You will design our data structures, configure our servers, and guide our MERN team on how to interface with the core search cluster correctly. Your Mission: To design and deploy a high-speed, enterprise-grade hybrid search engine (Lexical + AI Vector Semantic Search) that processes complex geographic filters and dynamic ranking rules in under 50 milliseconds across 5 million rows. Key Responsibilities: Cluster Mapping: Architect and configure our Elasticsearch instance, writing production-ready JSON mapping templates, synonym filters, and autocomplete completion suggesters. Geospatial Infrastructure: Set up our spatial database strategy using PostgreSQL/PostGIS. Integrate Uber’s H3 spatial index framework to handle high-speed radius lookups. AI Vector Engineering: Define the vector generation pipeline for unstructured text data and implement kNN (k-Nearest Neighbors) semantic searches fused with keyword queries via Reciprocal Rank Fusion (RRF). Algorithmic Ranking Logic: Write custom ⁠script_score⁠ structures inside Elasticsearch to combine text relevance with platform business rules (Bayesian review metrics, profile completeness scoring, and daily randomization fairness shuffles). Synchronization Strategy: Design a Change-Data-Capture (CDC) pipeline (e.g., using PGsync or equivalent WAL streaming loops) to batch-sync millions of relational database rows out to Elasticsearch without thread pool exhaustion. Team Guidance: Conduct architecture reviews and provide clear documentation/code boilerplates for our MERN developers so they can build standard API integrations flawlessly. Required Technical Stack Expertise: Search Engine: Elasticsearch / OpenSearch (Advanced level: Script scoring, Custom tokenizers, Hybrid search, RRF). Spatial Database: PostgreSQL with PostGIS extension (⁠ST_Contains⁠, ⁠ST_Distance⁠). Spatial Hashing: Uber H3 Grid System (Experience calculating expanding hexagon grid disks natively in-database or via Node.js). Geocoding Pipelines: Practical knowledge of open-source, self-hosted geocoding engines (Pelias or Nominatim/OpenStreetMap). AI/ML Integration: Vector embeddings generation (⁠text-embedding-3-small⁠ or HuggingFace open-source libraries) and vector search tuning.