Applied Researcher at Variant

Software should be beautiful. Variant is code generation with creativity and taste. Instead of a confined conversation, you can generate endless designs from a single idea. Freely explore, discover directions you wouldn’t have thought of, and build better software as a result. We're well-capitalized and looking for an applied researcher to join us at Variant. Job Responsibilities: Design, train, and evaluate deep neural networks and LLMs Own experiments end-to-end: hypotheses, datasets, metrics, results Prototype fast; turn promising ideas into reliable systems Collaborate with engineering and design to make ideas real Write clearly: papers, docs, and crisp experiment reports Don't compromise between rigor and shipping fast See what needs to be done and do it Minimal qualifications: 2+ years in PhD program or in industry Experience training deep neural networks Proficiency in frameworks like Pytorch Published 1 or more papers in the subject of deep learning Ideal qualifications: Experience training LLMs, including SFT, RL-based techniques, etc. Experience in the domain of code generation At least 1 paper accepted into a top conference Proficiency in best software engineering practi