IBM Research takes responsibility for technology and its role in society. Working in IBM Research means you'll join a team who invent what's next in computing, always choosing the big, urgent and mind-bending work that endures and shapes generations. Our passion for discovery, and excitement for defining the future of tech, is what builds our strong culture around solving problems for clients and seeing the real world impact that you can make.
IBM's product and technology landscape includes Research, Software, and Infrastructure. Entering this domain positions you at the heart of IBM, where growth and innovation thrive.
Research Scientist position at IBM India Research Lab is a challenging, dynamic and highly innovative role, where you will be responsible for coming up with new innovative ideas, developing solutions working as a team, building prototypes, publishing research papers and demonstrating the value of your ideas in an enterprise setting.
Some of our current areas of work where we are actively looking for top researchers are:
- Optimized runtime stacks for foundation model workloads including fine-tuning, inference serving and large-scale data engineering, with a focus on multi-stage tuning including reinforcement learning, inference-time compute, and data preparation needs for complex AI systems.
- Optimizing models to run on multiple accelerators including IBM’s AIU accelerator leveraging compiler optimizations, specialized kernels, libraries and tools.
- Innovative use cases that effectively leverage the infrastructure and models to deliver value
- Pre-training language and multi-modal foundation models working with large scale distributed training procedures, model alignment, and creating specialized pipelines for various tasks including effective LLM-generated data pipelines.
You should have one or more of the following:
- A master’s degree in computer science, AI or related fields from a top institution
- 0-8 years of experience working with modern ML techniques including but not limited to model architectures, data processing, fine-tuning techniques, reinforcement learning, distributed training, inference optimizations
- Experience with big data platforms like Ray and Spark
- Experience working with Pytorch FSDP and HuggingFace libraries
- Programming experience in one of the following: Python, web development technologies
- Growth mindset and a pragmatic attitude
- Peer-reviewed research at top machine learning or systems conferences
- Experience working with pytorch.compile, CUDA, triton kernels, GPU scheduling, memory management
- Experience working with open-source communities