About the role
As a Research Engineer at Gamma, you'll build models for visual communication, a foundational bet for the company. You'll teach models to reason about spatial composition, hierarchy, and visual language the way a skilled communicator or designer does.
This work sits at the intersection of research rigor and product impact. You’ll have the opportunity to build evals and training data for a field where there isn’t much of either. You’ll fine-tune vision-language models so that Gamma's 100M+ users get exceptional design every time they generate.
You'll succeed here if you combine deep expertise in VLMs and multimodal modeling with a research mindset, comfort working in ambiguity, and a rigorous eye for visual and design quality.
Our team has a strong in-office culture and works in person 4 to 5 days per week in San Francisco. We love working together to stay creative and connected, with flexibility to work from home when focus matters most.
What you'll do
Fine-tune vision-language models to generate and critique layouts, reason about design intent, and translate content into coherent visual form
Design evaluation frameworks and benchmarks for visual communication quality, covering layout, typographic structure, color, and information density, the dimensions generic text evals miss
Lead proprietary data collection for visual design tasks, building the datasets needed to teach models design principles that aren't available off the shelf
Run rigorous experiments to understand model behavior, then turn the results into targeted improvements: a new training objective, a fine-tuned model, or a sharper benchmark
Diagnose systematic failure modes in production output and fix them at the root rather than patching symptoms
Build the tools and workflows that let the team iterate and validate fast
Partner with product and engineering to ship quality improvements that hold up at scale
What you'll bring
Hands-on experience with vision-language models or multimodal modeling: training, fine-tuning, or systematically evaluating them
Experience with post-training techniques including supervised fine-tuning and reinforcement learning
Track record of building evaluations for subjective or hard-to-measure qualities, not just accuracy on labeled benchmarks
2+ years building AI systems, with production experience shipping models that real users depend on
Master’s or PhD in Computer Science, Machine Learning, or a related field, or equivalent hands-on research experience. A strong publication record at top-tier conferences such as NeurIPS, CVPR, ACL, or comparable venues.
Compensation range:
The base salary for this full-time position, which spans multiple internal levels depending on qualifications, ranges between $180K - $340K plus benefits & equity.
Final offer amounts are determined by multiple factors, including but not limited to experience and expertise in the requirements listed above.
If you're interested in this role but you don't meet every requirement, we encourage you to apply anyway! We're always excited about meeting great people.