Sergio Echeverri

Architected a retrieval-augmented generation pipeline connecting a Shopify product catalog to OpenAI embeddings, reducing repetitive customer-service ticket volume by ~22%. Built bidirectional synchronization between HubSpot CRM and the e-commerce order database using scheduled Python jobs and REST webhooks. Implemented a pgvector-backed vector database layer in Postgres for storing product/policy embeddings and powering order-status answers without human escalation. Diagnosed and resolved a production incident caused by malformed payment-processor webhook payloads by adding schema validation and alerting (detection ~20 minutes). Automated marketing funnel segmentation using Klaviyo and purchase-history scoring, improving click-through rate by ~15%. Consolidated fragmented spreadsheets and a legacy CRM export into a unified PostgreSQL schema for reporting, and normalized inconsistent supplier product metadata via Python ETL into BigQuery. Evaluated integration designs from junior contr

Developed Python/Flask RESTful APIs to serve order, inventory, and customer data to an e-commerce storefront and internal CRM. Integrated Zendesk with Mailchimp so support-ticket-history flags triggered targeted retention email sequences. Built a PostgreSQL data warehouse with star-schema modeling consolidating order, marketing, and support records, reducing engineering ad hoc reporting requests by ~40%. Supported faster campaign targeting with a lightweight logistic regression model trained on purchase history for a loyalty funnel. Migrated an on-prem order database to AWS (RDS plus scheduled Lambda jobs) for nightly CRM synchronization. Addressed rising abandoned support tickets by building a rules-based chatbot using spaCy intent classification, reducing backlog by ~18%. Wrote pytest unit/integration tests for CRM and marketing-automation connectors, improving regression prevention, and piloted GPT-3 integrations to draft agent reply suggestions, cutting reply composition time by ~2

Implemented backend order-processing endpoints in Python and Django for an online retail platform, covering cart, checkout, and payment-status updates. Designed normalized MySQL schemas and indexes to reduce slow-query timeouts during peak sale traffic. Debugged intermittent checkout failures caused by a race condition in inventory decrementing and added database-level locking to keep stock counts accurate. Automated nightly CSV export scripts feeding marketing email campaigns, reducing manual export effort by ~5 hours/week. Added unit/integration tests with JUnit to prevent regressions before staging, and supported production releases by triaging customer-reported bugs, reproducing issues in a staging clone, and coordinating fixes before deploys.