David Tan

Designed a phased architecture for an AI-enabled operations platform by assessing fragmented spreadsheets, Shopify exports, and support-ticket workflows. Built a retrieval-augmented generation pipeline over historical support tickets and product documentation using OpenAI embeddings and a Pinecone vector store to power automated customer-service responses. Implemented CRM and customer-support synchronization via custom webhooks and a Python service to update order status and customer profile details for ~15,000 monthly interactions. Reduced first-response time for routine inquiries from 6 hours to under 45 minutes by deploying an LLM-based triage and auto-response layer. Analyzed order/returns/churn data to inform marketing funnel updates. Implemented a lightweight automation stack using Make and serverless functions, reducing integration build time by ~40%. Reviewed prompt templates and automation workflows built by contract developers, establishing testing and rollback conventions pr

Built REST APIs connecting an order-management system to a third-party CRM so customer-service agents could access purchase history without switching tools. Developed Python and SQL ETL pipelines to consolidate marketing campaign data, support tickets, and sales records into a reporting warehouse used for weekly funnel analysis. Debugged recurring data-sync failures causing duplicate tickets by tracing issues to a race condition in the webhook handler. Automated recurring customer segmentation reports using scheduled Python scripts, reducing manual spreadsheet work from two days to about three hours weekly. Integrated a chatbot vendor’s API into support workflows to handle tier-1 inquiries and reduce ticket volume reaching human agents by ~25%. Migrated legacy customer records to cloud-hosted PostgreSQL while preserving referential integrity across order, inventory, and support tables. Collaborated with marketing on event-tracking schemas feeding abandoned-cart and re-engagement emai

Implemented Python and Flask backend endpoints for an internal customer-order lookup tool used by support teams. Wrote unit and integration tests for order-processing logic, increasing checkout module test coverage from ~50% to ~80%. Debugged intermittent payment-gateway timeouts by adding retry logic and improving error logging to prevent repeat failures. Built SQL reports for operations to track daily order volume, refund rates, and inventory discrepancies across two warehouses. Automated a manual nightly data-export process using cron-scheduled scripts, removing the task from the on-call engineer’s checklist. Supported production incidents during peak sales by monitoring application logs and escalating database performance issues to senior engineers.