About the role
About Ideogram 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 including early research in Diffusion Models, Google’s Imagen, and Imagen Video. We care about design, taste, and craft as much as research and engineering – shipping experiences that creatives actually love. We’ve raised nearly $100M, led by Andreessen Horowitz and Index Ventures. Headquartered in Toronto with a growing team in NYC, we're scaling fast, aiming to triple over the next year. We're a flat team with a culture of high ownership, collaboration, and mentorship. Explore Ideogram 3.0 , Canvas , and Character blog posts, and try Ideogram at ideogram.ai . The Opportunity In this role, you will develop the post-training pipeline for our text-to-image foundation models end to end, from data strategy to deployment, advancing techniques such as RLHF, RLAIF, and work on personalization/customization. You will contribute to post-training research that drives measurable gains, and implement and maintain high-throughput fine-tune/eval pipelines. You'll work with a creative and ambitious team of engineers and researchers who are building the future of the creative economy.
What We're Looking For 5+ years of experience in developing machine learning models in JAX, PyTorch, or TensorFlow. Experience in implementing Machine Learning foundations ( e.g., Transformer , VAE , Denoising Diffusion models ) from scratch. Track record in machine learning innovation and familiarity with Deep Learning and advanced Machine Learning. End-to-end understanding of generative media applications and excitement for pushing the state-of-the-art in generative AI. Ability to debug machine learning models to iteratively improve model quality and performance.