Latest AI JobsBrowse Jobs
Sarvam AI logo

ML Researcher, Foundational Models

Sarvam AIBengaluruModels
Apply on company site
Salary
Not listed
Job type
Full-time
Posted
21 May 2026
Status
Open · checked 7d ago

About this role

Role Overview

Sarvam AI is hiring an ML Researcher, Foundational Models in Bengaluru to shape the architecture and training strategies for next-generation AI systems. This position suits an experienced researcher who wants to drive core investigations into scaling behavior, optimization, and foundational model development.

Key Responsibilities

Lead open-ended exploratory research focused on architecture, optimization, scaling behavior, training stability, and post-training methods for future foundational models.
Create and execute ablations at a scale capable of influencing large production training runs, including managing pre-training experiments independently from start to finish.
Convert research insights into actionable proposals for upcoming training runs and oversee their execution into production environments.
Collaborate closely with infrastructure and data engineering teams to investigate research questions situated at system boundaries.
Stay current with academic literature, write internal documentation, and share findings externally through publications when appropriate.

Requirements

Hold or be in the final stages of completing a PhD in Computer Science, Machine Learning, or a closely related academic discipline.
Bring at least three years of post-PhD research experience or equivalent depth, with strong consideration given to exceptional early-career applicants.
Demonstrate a history of first-author publications at premier machine learning conferences such as NeurIPS, ICML, ICLR, ACL, EMNLP, or COLM.
Possess practical, hands-on experience pre-training transformer-based language models from scratch, preferably at a scale of 7B+ parameters.
Show active engagement with the open-source large language model community through research code, model releases, datasets, or project contributions.
Exhibit high fluency in PyTorch alongside deep familiarity with distributed training environments and the ability to diagnose training loop bottlenecks or instability.
Demonstrate strong experimental intuition regarding proper measurement, ablation design, and the scale required to trust results.

Nice to Have

Experience investigating novel architectures like mixture-of-experts, state-space models, hybrid designs, or custom modifications to standard transformers.
Background working with multilingual or multimodal pre-training workflows.
Prior research contributions within post-training domains including reinforcement learning from human feedback, distillation, or reasoning models.
A proven history of advancing research concepts from initial prototypes into deployed capabilities.

Skills and Tools

PyTorch
Large Language Models (LLMs)
Distributed training frameworks
Transformer architectures

About the Company

Sarvam AI is developing India's comprehensive sovereign AI platform by innovating across research, infrastructure, models, and applications to make artificial intelligence accessible for the region. The organization collaborates with major enterprises and public institutions, partnering with leading brands like Tata Capital, SBI Life, CRED, IDFC, and LIC. Backed by prominent investors including Lightspeed, Peak XV, and Khosla Ventures, Sarvam maintains a fast-moving, high-talent environment focused on population-scale impact.

How to Apply

Interested candidates should visit the official Sarvam AI careers page online to submit their application for this Bengaluru-based opportunity.

Skills and tech stack

PyTorchLLM

Questions candidates ask

Where is this role located?

This position is based in Bengaluru, India.

What level of experience is required?

The role requires a PhD in Machine Learning, Computer Science, or a related field, along with 3+ years of post-PhD research experience or equivalent depth.

Is this a fine-tuning role?

No, this is not a fine-tuning or recipe-following position; it involves open-ended foundational model research, architecture design, and pre-training experimentation.

What technical framework fluency is expected?

Candidates must be fluent in PyTorch and comfortable working with distributed training setups.

Who backs Sarvam AI?

Sarvam AI is backed by Lightspeed, Peak XV, and Khosla Ventures.

Ready to apply?

The application is on the official Sarvam AI careers page.

Apply on company site