Agents and Robots: The Human Effort Putting AI Into Production
summary
Encord ingests multimodal logs (LiDAR, depth, teleop, video) from cloud storage, searches at scale for long‑tail events, and applies model‑led prelabeling (Invideo, OpenAI, Anthropic) before human review. Teams combine automated labeling and QA, track every model and human action for auditability, and measure annotator quality. Customer results show fewer annotation tasks, higher MAP, and 60% faster model iteration after deploying pre-labeling with OpenAI/Anthropic/Invideo and structured human handoffs.
AI-written summary · about
related talks
17mYour AI Deserves Better Data
37mBest Practices for Building AI Datasets
15mChallenges in moving Gen AI apps to production
18mRobotics: why now?
46mFullstack AI: The Building Blocks for Production Ready AI-Powered Software
28mAgentic AI
25mMonitoring AI with AI
53mArtificial vs Natural Intelligence in Annotation Tasks