ML Engineer – Inference/Training Optimization

As a Machine Learning Engineer at Krea, your key responsibilities include writing custom CUDA Kernels to improve the speed of multi-node inference for image and video models, while also working on caching mechanisms and dynamic compilation techniques. You'll be tasked with optimizing the efficiency of training runs across our powerful GPU clusters, making your role essential in enhancing the performance of our AI offerings.

To be considered for the Machine Learning Engineer position at Krea, candidates should possess expertise in CUDA or parallel programming and have a strong background in Python or C++. Experience in optimizing diffusion and transformer models for both performance and scalability is critical. Additionally, candidates should exhibit high agency and resourcefulness, indicating a strong ability to solve complex problems independently.

In the role of Machine Learning Engineer at Krea, you will directly contribute to the enhancement of AI models by implementing optimization techniques. This includes writing efficient CUDA Kernels and exploring innovative caching and dynamic compilation strategies. Your work will ensure faster inference and more efficient training processes, significantly elevating the capabilities of the models we develop.

A Machine Learning Engineer at Krea will engage in exciting projects that revolve around optimizing models for various formats such as text, image, and video. You will work closely with AI researchers and infrastructure teams to integrate optimizations into existing frameworks, effectively working on cutting-edge AI applications that push the boundaries of creativity.

At Krea, the work environment for a Machine Learning Engineer is collaborative and innovative, located in the tech hub of San Francisco. You will be part of a team that values creativity and the power of AI, encouraging you to experiment, share ideas, and contribute to meaningful projects. The culture fosters resourcefulness and independence, making it an ideal setting for passionate individuals.

When addressing your experience with CUDA programming, be sure to highlight specific projects where you developed custom kernels. Discuss the impact your work had on performance and any challenges you faced and overcame. Sharing a clear example that demonstrates your skills can significantly strengthen your response.

In your answer, categorize the techniques you typically use, such as reducing model size, employing quantization, or implementing caching strategies. Also, discuss your experience with transformer models and how you’ve implemented performance optimizations in previous projects. Clear, specific examples will showcase your expertise.

Discuss your systematic approach to debugging GPU-accelerated applications, mentioning tools like NVIDIA Nsight or CUDA-GDB. Highlight how you identify bottlenecks and optimize code iteratively. Providing a past experience where you diagnosed and solved a major issue can offer valuable context.

Talk about the frameworks and methods you've used for multi-node training, such as distributed training with TensorFlow or PyTorch. Include specific details on how you ensured synchronization and data sharing across nodes, showcasing your understanding of distributed systems.

Provide a detailed description of a specific project where your optimizations led to measurable improvements. Include the techniques you used (like caching or algorithmic changes) and the results in terms of latency reduction, emphasizing your contributions to the project's success.

Share your strategies for data management, such as versioning, preprocessing steps, and data augmentation techniques. Discuss how these strategies support efficient loading and processing, particularly in training and inference contexts, demonstrating a comprehensive understanding of data handling.

Mention how you actively engage with the AI community through research publications, forums, blogs, and conferences. Discuss any specific influential papers or technologies you've learned from, showcasing a commitment to continuous learning in a rapidly evolving field.

In your response, detail your collaborative experiences, emphasizing communication, brainstorming innovative solutions, and implementing research findings into practical applications. Highlight any successful projects where your contributions made a tangible impact.

Discuss your strategies for prioritization, such as using project management tools, understanding project timelines, and defining key milestones. Demonstrating your organizational skills and ability to adapt can convey your aptitude for thriving in a dynamic work environment.

Express enthusiasm for the transformative power of AI and its ability to enhance human creativity across different mediums. Highlight your vision of future applications and the importance of ethical considerations in AI development, reflecting your passion and alignment with Krea's mission.