Join Sarvam in developing India's sovereign AI platform. This role focuses on foundational model research, pushing the boundaries of AI architecture, optimization, and training dynamics. You'll have direct access to substantial compute resources and collaborate closely with engineering teams to drive innovation.

This is a unique opportunity to shape the future of large-scale AI models with direct impact on India's technological landscape. You'll be part of a fast-paced, high-talent team committed to building AI solutions for India.

Drive pioneering research into AI model architectures, optimization techniques, scaling behaviors, and training stability. Design and execute impactful ablation studies at scale to inform critical decisions for production training runs.

Translate research breakthroughs into actionable proposals for upcoming training cycles, overseeing their implementation and successful integration. Collaborate extensively with infrastructure and data pipeline teams, tackling complex research challenges at their intersection.

A PhD in Machine Learning, Computer Science, or a related field is required, or you should be in the final stages of completion. Candidates with exceptional research records and equivalent depth of experience will be considered.

Possess a minimum of 3 years of post-PhD research experience. Demonstrate a strong publication record in top-tier ML venues such as NeurIPS, ICML, ICLR, ACL, EMNLP, or COLM (first-author publications are highly valued).

Exhibit hands-on expertise in pre-training transformer-based language models from scratch, ideally exceeding 7 billion parameters. Showcase a deep understanding of experimental design, fluency in PyTorch, and proficiency in distributed training environments. Experience debugging complex training loops is essential.