Project Hedgehog Multimodal AI Trainer (Handshake AI)
Don T DiscloseAudioAudio RecordingTranscription
Project Hedgehog is a large-scale multimodal alignment initiative focused on training frontier Large Language Models to interpret and generate content across text, audio, and video domains. As a specialized trainer, I performed high-precision data labeling tasks including audio data synthesis, image bounding box annotation, and video-to-text prompt engineering to improve the model’s spatial and auditory awareness. I was responsible for developing and iteratively editing complex prompts to test the model's logical reasoning, while adhering to rigorous quality benchmarks and technical rubrics to ensure the output was factual, safe, and instruction-compliant.
Project Hedgehog is a large-scale multimodal alignment initiative focused on training frontier Large Language Models to interpret and generate content across text, audio, and video domains. As a specialized trainer, I performed high-precision data labeling tasks including audio data synthesis, image bounding box annotation, and video-to-text prompt engineering to improve the model’s spatial and auditory awareness. I was responsible for developing and iteratively editing complex prompts to test the model's logical reasoning, while adhering to rigorous quality benchmarks and technical rubrics to ensure the output was factual, safe, and instruction-compliant.
2025 - 2025