Applied Scientist Intern

Summary

Join Reality Defender, a Y Combinator graduate and Comcast NBCUniversal LIFT Labs alumni, as we pioneer multi-modal and multi-model detection of AI-generated media. We offer a research-forward environment focused on combating fraud and disinformation in real time. This role involves exploring novel methods for deepfake detection, conducting cutting-edge research in audio deepfake detection, building generalizable models, and collaborating with scientists and engineers. You will also be responsible for summarizing, publishing, and presenting research findings. This position requires a PhD program enrollment in a relevant field, significant experience with large model training, and a strong publication record.

Requirements

• Currently enrolled in a PhD program with specialization in machine learning/deep learning, natural language processing, and/or speech processing
• 2+ years of experience with training/fine-tuning large models, esp. audio language models, speech foundation models, multi-modal foundation models
• Experience with end-to-end model building pipeline for ML tasks: dataset curation/cleaning, model implementation, benchmarking, and result analysis
• Familiarity with distributed multi-GPU training
• Prior experience with publications in reputable ML/Audio/NLP research venues, e.g. NeurIPS, Interspeech, ICASSP, ACL, EMNLP

Responsibilities

• Explore and conceptualize novel methods to leverage different modalities (e.g., speech, text) for deepfake detection and relevant audio understanding tasks
• Perform fundamental and applied research to advance the current state-of-the-art on audio deepfake detection
• Build models with generalizability to unseen generative methods
• Collaborate with scientists and engineers across the organization
• Summarize, publish, and present research findings