Adrián Carbonell Sánchez

Designed and integrated multi-agent LLM workflows into backend services; built vector-database retrieval pipelines to support production RAG features; created AWS-based ingestion and indexing pipelines for documents and embeddings; collaborated with frontend and DevOps to deploy AI services into the product; recruited into the role following a prior team layoff.

Designed and integrated multi-agent LLM workflows into backend services using Python and Go. Implemented vector database retriever pipelines to support production RAG features. Built AWS-based ingestion pipelines for document indexing and embedding generation. Collaborated with frontend and DevOps teams to deploy and integrate AI services into the product. Recruited by my previous manager following mass layoffs.

Integrated an LLM-powered voice assistant with Retrieval-Augmented Generation (RAG) into the vehicle backend, enabling natural-language vehicle commands; deployed to ~90% of the fleet with ~40% 30-day retention. Built a low-latency streaming pipeline: wake-word → ASR → LLM response → TTS; end-to-end p95 latency: 2s. Implemented RAG using Azure AI Search (vector embeddings) and backend orchestration, including intent detection and safety validation. Instrumented OpenTelemetry for usage, uptime, and LLM/MLOps metrics; dashboards and alerts. Tech: Azure OpenAI, Azure Speech (ASR/TTS), Azure AI Search, Golang, Docker, Terraform, OpenTelemetry, Grafana.