Internship - MScAC 2027
Core
Build pipelines, benchmarks, and models for real-time speech, multimodal AI, and voice agents to make talking to AI feel natural.
Role type
MScAC applied research intern (audio, video, systems, agents)
Builds
Speech-to-speech API, text-to-speech models, talking avatars, and voice agents
Domain
AI/ML, speech processing, computer vision, real-time systems
Deliverable
production ML models
Required skills
Machine learning, statistics, Python, CUDA/Triton, diffusion models, reinforcement learning, systems optimization
Preferred skills
Multilingual ability, experience with dialog systems, human-computer interaction, image/video generation
Technologies
PyTorch, CUDA, Triton, SGLang, vLLM, diffusion models, Transformers
Responsibilities
Build data pipelines for audio collection and annotation; design benchmarks for expressive speech and voice cloning; develop audio-driven video generation and lip-sync models; optimize training on 900-GPU clusters; improve real-time inference speed and cost; design conversational agents with theory of mind; work on full-duplex audio and audio-video models
Seniority
Intern
