The real jobYou'll be teaching computers to understand human behavior through video. Not just "is there a face?" but "is this person confident or just good at faking it?" This is computer vision meets psychology meets real-time processing at scale.

Your actual work will involve:

• Building systems that analyze body language, facial expressions, eye contact during interviews

• Creating models that work across different lighting, cameras, and network conditions (spoiler: everyone's webcam sucks)

• Detecting if someone's reading answers off screen (harder than you think)

• Building privacy-preserving vision systems (we analyze but don't store faces)

• Optimizing models to run real-time on potato-quality video streams

• Fighting with WebRTC, cursing at codec issues, and somehow making it all work

You probably have:

• Built CV systems that work outside of perfect lab conditions

• Experience with real-time video processing (not just image classification)

• Deep knowledge of modern CV architectures (YOLO, Vision Transformers, etc.)

• Understanding of edge deployment and model optimization

• Battle scars from deploying CV in production

• Projects involving human behavior analysis or video understanding

Epic work we're looking for:

• Your AR filter went viral

• You've built CV systems processing millions of images/videos daily

• You've published CV research that people actually implemented

• Your open-source CV project has thousands of stars

• You've built something like Snapchat filters but better

• You made a CV model run on a Raspberry Pi doing something actually useful

Why people fail:They've only worked with static images. Or they think accuracy is the only metric that matters. Or they can't handle the mess of real-world video data.