Handling Offline AI Jobs: Developers Share Best Engineering Practices
cmm324 · reddit · 2026-08-03
A developer started a Reddit discussion asking the community about their current engineering approaches to handling offline/batch inference workloads that don't require immediate responses.
The author listed common asynchronous workloads, including:
- Generating embeddings for large corpora
- Enriching product catalogs
- Evaluating prompts or models
- Image or video generation queues
- Processing millions of rows
- Nightly or weekly pipelines
They specifically asked for real-world experiences regarding solutions like Provider batch APIs, custom queues and workers, Airflow/Temporal/Celery, Kubernetes Jobs, and AWS Batch/Spot. What works well? What was more painful than expected? And if they could change one thing about their current setup, what would it be?
More from coding & agent
- OpenAI Codex Automates Ad Campaigns End-to-End — gdb · 2026-08-03
- Dev Builds Ambient-Aware AI App: Global Agents with Screen Context & Visual Particles — RileyRalmuto · 2026-08-03
- ADHD Skill: Adding Tree-of-Thought Pruning to AI Coding Agents — Saboo_Shubham_ · 2026-08-03
- Doppel: A Low-Token MCP Server for Real Browser Automation — No_Airport_1450 · 2026-08-03
- Boosting Codex Efficiency: Stacking More Skills Per Thread — EverydayAI_ · 2026-08-03
- Automate Codex Cleanup Scripts to Cool Down Your Computer and Keep It Smooth — alexcovo_eth · 2026-08-03