AI
UV Engine
A background job processor for handling multiple AI workloads — built with Redis, BullMQ, and the Gemini API, with an evaluation engine to verify outputs.
Overview
UV Engine is a processor built to handle multiple jobs in the background using Redis and BullMQ. The project started as a deep dive into async workflows — how to queue work, process it reliably, and keep the main thread free while AI tasks run. It integrates the Gemini API for generation and includes an evaluation engine that checks whether model output is correct before it moves forward.
Features
- —Background job queue with Redis and BullMQ
- —Multi-job processing for concurrent AI workloads
- —Gemini API integration for generation tasks
- —Evaluation engine to verify AI output quality
- —RAG pipeline (in progress — targeting next weekend)
- —More reliability and observability improvements planned
Tech Stack
RedisBullMQGemini APINode.jsTypeScriptAI Evaluation
What I Learned
- —How background job queues work with Redis and BullMQ
- —Designing async workflows that don't block the main process
- —Integrating the Gemini API into a production-style pipeline
- —Building an evaluation layer to validate LLM output
- —Why verification matters before trusting AI-generated results