Research Engineer/Research Scientist (Post-training)

Responsibilities

• Develop the end-to-end post-training pipeline for text-to-image foundation models, from data strategy to deployment
• Advance techniques such as RLHF, RLAIF, and personalization/customization
• Contribute to post-training research that drives measurable gains in model quality
• Implement and maintain high-throughput fine-tuning and evaluation pipelines

Requirements

• 5+ years of experience in developing machine learning models using JAX, PyTorch, or TensorFlow
• Proven experience implementing ML foundations (e.g., Transformers, VAE, or Denoising Diffusion models) from scratch
• Deep understanding of generative media applications and the ability to push the state-of-the-art in generative AI
• Strong ability to debug machine learning models to iteratively improve performance and quality

Preferred Qualifications

• Familiarity with Kubernetes and Docker
• Experience in low-level machine learning optimization, such as writing CUDA kernel code

Benefits

• Competitive compensation and equity
• 4 weeks of vacation
• Comprehensive health, vision, and dental coverage starting on day one
• RRSP/401(k) with employer match up to 4%
• Top-of-the-line tools and tech
• Toronto HQ perks including daily in-office lunches and dinners
• Autonomy to explore, experiment, and access necessary compute resources

About the Company

Ideogram's mission is to make world-class design accessible to everyone, multiplying human creativity. We build proprietary generative media models and AI-native creative workflows, tackling unsolved challenges in graphic design. Our team includes builders with a track record of technology breakthroughs in Diffusion Models and Google’s Imagen. We are a fast-scaling team with a culture of high ownership, collaboration, and mentorship.