Mynd merged into Roofstock

• Built a centralized property intelligence platform aggregating 130M+ U.S. single-family homes (183 attributes, 13 data sources), served via gRPC/REST APIs on Snowflake Hybrid Tables. Cut property evaluation time from ~4 hours to ~1 hour, enabling analysts to process 4× more properties daily.

• Designed a field-level source priority framework to automatically resolve conflicts across overlapping data providers — each attribute governed by its most authoritative source without manual reconciliation.

• Engineered an event-driven enrichment pipeline (flood zones, school scores, AVMs, census, CBSA) using SQS to decouple 6 external enrichment services from the fast write path, keeping API latency within budget.

• Built a global property identifier (GRID) service to normalize and deduplicate addresses across 13 sources using 2-tier normalization (SmartyStreets → internal parser), deterministic hashing, and DB-level uniqueness constraints to guarantee zero duplicates under concurrent writes. Implemented multi-level caching that cut third-party address validation costs by ~5× (450M → 75M annual calls).

• Designed and built the Service Request platform from scratch — the core repair/maintenance system for 40K+ properties. Built as a CQRS event-driven modular monolith (430+ command/event handlers, 40+ gRPC integrations), automating ~75% of workflows with multi-level approvals, vendor auto-dispatch, and SLA tracking.

• Built the Vendor Portal API (100+ endpoints) for 2,000+ vendors to manage end-to-end repair workflows. Reduced vendor onboarding time by 3× and job completion time by 25%.

Integrated OpenAI into the Service Request pipeline to auto-summarize work orders, validate resident-submitted requests, and verify vendor-reported work — reducing manual triage and QA effort across both sides.

• Built a multi-channel communication service processing 1M+ messages/month (SendGrid, Twilio, push, Slack) with event-driven async pipelines on DynamoDB and SQS.

• Architected a company-wide Geo API, migrating legacy geography models to U.S. Census Regions and Divisions. Led cross-team coordination and migration planning, reducing reporting prep time by 20%.

Mentored engineers, established SDLC practices, and improved delivery predictability from 65% to 85%, reducing time-to-market by 20%.

• Built a high-throughput messaging system (~20K orders/hour, millions of messages/day) for Ocado,

the UK's largest online-only grocer. Implemented bi-directional integration connecting warehouse

management to multiple e-commerce platforms using Kafka, Cassandra, and Apache Camel.

• Developed backend services and APIs for an Analytical Reporting Enterprise System for a major

Wall Street client, powering a spreadsheet-based interface that enabled analysts to query and

aggregate data from heterogeneous sources (SQL databases, external services) in real-time.