IBM Research is the innovation and growth engine of the IBM corporation. It is the largest industrial research organization in the world with 12 labs on 6 continents. IBM Research produces more breakthroughs–more than 9 patents are produced every day–than any other organization in the world. IBM employs over 3200 researchers worldwide.
IBM Research India (IRL) is the leading industrial research lab in India, shaping the future of computing across AI, Hybrid Cloud and Quantum Computing. IRL has a long legacy of ground-breaking innovation in the areas of computer science and its applications to a wide variety of disciplines and offerings for IBM. IRL researchers are working on projects that are pushing the state of the art across Foundation Models, optimized runtime stacks for FM workloads such as tuning, large scale data engineering and pre-training, multi-accelerator model optimization, agentic workflows and modalities across language, code, time series, IT automation and geospatial. We are strong proponents of open-source community-driven software and model development, and our work spans a wide spectrum from research collaborations with academia to developing enterprise-grade commercial software.
Research Engineer position at IBM India Research Lab is a challenging, dynamic and highly innovative role.
We are actively looking for top talent in the area of:
- Software stack optimization for IBM’s Spyre accelerator, including compiler enhancements, specialized kernels, performance libraries, and tooling.
- Low-level optimization within the PyTorch stack or below, aimed at maximizing GPU resource utilization.
You should have one or more of the following:
- A Master’s/PhD degree in Computer Science, AI, or a related field from a top-tier institution
- 0–8 years of experience in the Systems for AIdomain, with expertise in one or more of the following:
- Model architectures
- Distributed training
- Inference optimization
- GPU or other accelerator architectures
- Multi-accelerator networking (e.g., NCCL)
- Compilers
- CUDA programming
- Triton kernel development
- Experience with PyTorch FSDP and HuggingFace libraries
- Proficiency in Python or C++
- A growth mindset and a pragmatic, problem-solving attitude