SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mirage is an AI-native video platform that uses natural language to orchestrate production and editing, enabling professional-grade video creation at scale. The company has raised $75M from top-tier investors including Sequoia, Andreessen Horowitz, and Kleiner Perkins, and was recognized on Forbes' AI 50 list.
As a Research Engineer on the Generative Video team, you'll work at the intersection of research and systems engineering to build and scale the models powering Mirage's video generation capabilities. This role focuses on translating cutting-edge research into production systems that are faster, more efficient, and capable of ultra-low latency, real-time generation.
Key responsibilities include:
- Training and optimizing large-scale video and multimodal models
- Improving efficiency across training and inference pipelines (memory, latency, cost)
- Implementing advanced techniques such as distillation, quantization, and pruning to accelerate diffusion and autoregressive generation
- Building and maintaining distributed training systems with optimized GPU utilization and parallelism
- Developing tooling for experimentation, evaluation, and debugging
- Translating research models into robust, production-ready systems
- Monitoring and improving model performance in real-world usage
You'll need a BS/MS/PhD in CS, ML, or related field with 2+ years of professional industry experience. Strong expertise required in deep learning systems infrastructure, PyTorch, CUDA, Triton, and distributed training frameworks (FSDP, etc.). Experience scaling and optimizing large models under low-latency inference constraints is essential, along with strong debugging and performance profiling skills. The ideal candidate can move quickly from prototype to production.
This is a full-time, in-person role based at Mirage's headquarters in Union Square, NYC.
