John Terry

Built backend services integrating Amazon Bedrock model APIs across summarization, question answering, and document processing workflows. Developed retrieval-augmented generation pipelines with grounded context, improving p95 retrieval latency from 800ms to 520ms. Created reusable model-integration components for prompt construction, response parsing, request validation, and robust error handling. Improved reliability with timeouts, bounded retries, concurrency controls, and fallback behavior for model endpoint or downstream stream failures. Implemented authorization checks and scoped data retrieval to protect sensitive information across request lifecycle. Added automated evaluation workflows for answer relevance and factual grounding, reducing unsupported answers from 12% to 8%, and built operational dashboards tracking model errors, token usage, latency, and resource consumption.

Built and integrated backend services connecting Amazon Bedrock model APIs across multiple AI workflows, including summarization, question answering, and document processing. Developed RAG pipelines with source-grounded context, reducing p95 retrieval latency, and created reusable model-integration components for prompt construction, response parsing, request validation, and robust error handling. Added operational reliability features such as timeouts, bounded retries, concurrency controls, and fallback behavior for downstream/model unavailability. Implemented authorization checks and scoped retrieval to protect sensitive information, plus automated evaluation workflows for answer relevance and factual grounding with improved quality on a fixed dataset. Added operational metrics and dashboards for model errors, token usage, latency, and resource consumption. Also delivered enterprise AI workflows using Amazon Q Business and Amazon Q Developer, including assistant integrations for inf

Developed enterprise assistant integrations using Amazon Q Business, supporting information discovery for 500 monthly active users. Built document synchronization workflows for content and permission updates, reducing p95 sync lag from 60 minutes to 15 minutes. Integrated identity and document-permission mappings so retrieval results respected access restrictions from source systems. Developed API integrations for business actions with input validation, authorization checks, and recoverable error handling. Built test scenarios for retrieval relevance, source attribution, stale content, and unauthorized access to indexed information. Integrated Amazon Q Developer into multiple engineering workflows (code explanation, test development, and troubleshooting) to assist developers with engineering review feedback. Added monitoring for connector health, synchronization delays, ingestion errors, and assistant failures to support operational investigation and uptime.