About the Role
The role involves designing, training, and optimizing diffusion-based models for high-quality image generation and vision tasks. You will work on improving model architecture, scaling training pipelines, and integrating models into production systems.
Responsibilities
- Design and implement diffusion models for image generation
- Optimize training pipelines for large-scale vision datasets
- Collaborate on improving model fidelity and sampling efficiency
- Develop evaluation frameworks for generative outputs
- Integrate models into scalable inference systems
- Troubleshoot model convergence and stability issues
- Contribute to research on novel diffusion architectures
- Work with cross-functional teams to deploy vision models
- Monitor performance and accuracy in production environments
- Refactor code for maintainability and scalability
- Implement data augmentation strategies for training stability
- Analyze failure modes in generated outputs
- Support versioning and reproducibility of experiments
- Document model design decisions and results
- Stay current with advancements in generative modeling
- Participate in code reviews and technical discussions
- Tune hyperparameters for optimal model performance
- Assist in defining dataset curation standards
- Evaluate trade-offs between model size and inference speed
- Ensure ethical considerations in generative outputs
Nice to Have
- PhD in machine learning, computer vision, or related field
- Publications in top-tier AI conferences
- Contributions to open-source machine learning projects
- Experience with latent diffusion models
- Work on text-to-image generation systems
- Knowledge of attention mechanisms in vision models
- Experience with model distillation techniques
- Familiarity with reinforcement learning concepts
- Background in signal processing
- Prior industry experience in AI research
Compensation
Competitive salary with performance-based incentives
Work Arrangement
Hybrid work model with flexible remote options
Team
Collaborative AI research and engineering team focused on generative models
Research Focus
- Active exploration of next-generation diffusion techniques
- Focus on improving sample quality and diversity
- Investigation of multimodal conditioning strategies
Tech Stack
- PyTorch for model development
- CUDA for GPU acceleration
- Kubernetes for orchestration
- Weights & Biases for experiment tracking
Collaboration Model
- Weekly research syncs with engineering team
- Bi-weekly paper reading group
- Quarterly roadmap planning sessions
Available for qualified candidates
