Charlie Parrish

I’m a Senior AI Engineer who builds LLM-powered product systems end-to-end—from data ingestion and backend services to generation pipelines and evaluation loops. My focus is shipping AI features that are grounded in real user data and operate reliably at scale.

At Klaviyo, I built an LLM-driven campaign generation backend using Python microservices on AWS, enabling the K:AI Marketing Agent to auto-generate email/SMS campaigns, flows, and signup forms from website content. I reduced campaign setup time by ~70% and increased user adoption by ~30%.

I lead retrieval-augmented generation (RAG) pipelines and automated LLM evaluation and safety workflows. By incorporating client catalogs and website data into prompts and using monitoring to detect hallucinations and enforce brand voice/toxicity constraints, I reduced invalid or off-brand outputs by ~25–30% while improving content quality consistency.

I also optimize production AI for cost, latency, and reliability—implementing prompt compression, semantic caching, workload-based model routing, and asynchronous Celery/RabbitMQ workflows on Kubernetes. These efforts lowered per-campaign generation costs by ~35%, improved delivery automation (reducing manual marketer workload by ~50%), and helped sustain ~99.9% uptime.