Job Details

At Sieve, we're reimagining how machines understand video. We're a fast-growing team building cutting-edge AI systems that extract meaning from dynamic, real-world content-automatically, scalably, and in real time.
We're looking for a Applied Research Engineer who thrives in fast-paced, high-ownership environments and wants to push the boundaries of multimodal AI. You'll work with some of the latest Visual Language Models (VLMs) and generative models to build systems for a wide range of video understanding, editing, and search applications. This often means tackling ambiguous research problems and using creative approaches to solve them.

What You'll Do

• Design, train, and deploy deep learning models to extract insights from video at scale
• Build robust, performant pipelines in Python for real-time and batch processing
• Squeeze every drop of performance from models through clever pre/post-processing, pipelining, inference optimization, and fine-tuning
• Collaborate closely with product and customers to turn ambiguous problems into working systems
• Prototype with VLMs and multi-modal architectures to push Sieve's technology forward

What We're Looking For

• 4+ years experience in ML engineering, ideally in a startup or similarly scrappy, fast-moving environment
• Strong background in deep learning (PyTorch/TensorFlow) and computer vision
• Fluency in Python and best practices for scalable ML code
• Experience building AI systems to solve real customer problems
• A portfolio of side projects, demos, or contributions (GitHub, HuggingFace, etc.)
• Bonus: research background in ML, experience with CUDA, or open-source contributions

At Sieve, we care deeply about ownership, speed, and craft. If you're someone who loves building impactful ML systems, iterating quickly, and solving novel problems with creativity and rigor-you'll feel right at home.