Research Scientist - Generative AI

About Backflip
Backflip (backflip.ai) is a generative AI startup revolutionizing how engineers and designers create for the physical world. Founded by the team behind Markforged, we’re building AI-first tools that make 3D design as intuitive and fast as sketching on paper. Backed by a16z and NEA, we’ve raised a $30M Series A and built a deeply technical, mission-driven team of 15. Our goal is to empower one engineer to work at the speed of a hundred.

The Role
We are looking for a Research Scientist to design, train, and evaluate frontier generative models, uncovering ideas that work reliably at scale and translating them into production systems. You will own ambitious research directions from concept through prototype, working directly with our CTO and collaborating across the stack with our core engineering team on the company’s most important projects.

This is an opportunity to explore open-ended problems, develop novel architectures and evaluation methods, and deliver breakthroughs that unlock entirely new product capabilities. You will work in a fast-paced, high-ownership environment where shipping impactful ideas matters as much as publishing them.

We Expect You To

• Have a proven track record of introducing new ideas or improving existing ones in ML, demonstrated through first-author publications or impactful projects

• Bring direct, hands-on experience designing, training, and evaluating models, including developing novel measurement or mitigation techniques

• Independently define and pursue a research agenda, selecting high-impact problems and executing long-term projects

• Collaborate seamlessly with both research and product teams, with a strong bias toward impact and adaptability

• Step up to own challenges across the stack, even outside your formal role

Nice to Have

• Experience with high-performance, large-scale ML systems

• Expertise in designing, building, or deploying generative AI models

• Strong PyTorch skills and experience with performance optimization

• Experience working with large or complex datasets, including ETL pipelines

• A competitive, hacker mindset with the drive to win and ship results

Representative Projects

• Designing and training a new generative model architecture to capture complex geometric or structural patterns for new product capabilities

• Scaling distributed training to dramatically improve throughput and model quality in production

• Developing novel evaluation methods to assess generative model fidelity, diversity, and usability for real-world workflows

• Optimizing the compute efficiency of a new model variant for deployment in user-facing systems

• Creating interactive tools for visualizing and interpreting model behavior to guide product and research decisions

Why Backflip
Repeat founders in 3D printing/CAD with a prior $2.1B exit
Backed by a16z and NEA; small, senior team with outsized impact
The chance to help define the future of AI-powered design from the earliest stages