IBM expanded two Indian research collaborations on 1 October, extending work begun with IIT Bombay in 2018 and IISc in 2021 into agentic AI, sovereign AI and quantum computing, CXOtoday reported. The programmes cover separate research agendas, from adapting models for Indian languages to developing algorithms that combine quantum and high-performance computing.
Key points
- IBM and IIT Bombay plan research on Indic-language models, multimodal systems and enterprise knowledge retrieval.
- The IISc work includes agentic workflows across hybrid-cloud environments and Granite-based models for energy analytics.
- IBM and IISc will also explore quantum-HPC algorithms and the orchestration of quantum and classical systems.
IIT Bombay work targets Indian languages
IBM researchers will work with IIT Bombay faculty and students on adapting sovereign and Indic-language models. The planned work includes techniques intended to make model adaptation and optimisation more efficient for multilingual applications in Indian languages. It extends a research relationship that began in 2018.
The partners also plan to develop multimodal AI systems for software programming education, collaboration between people and AI, and operations across hybrid-cloud environments. Their infrastructure work will examine AI runtimes and distributed inference, with the aim of improving the performance and efficiency of large language models.
Enterprise knowledge retrieval is another part of the IIT Bombay programme. IBM and the institute aim to develop systems that give organisations faster and more accurate access to their information at scale. That work sits alongside the proposed changes to the computing systems used to run and serve models.
Ganesh Ramakrishnan, an Institute Chair Professor in IIT Bombay’s Department of Computer Science and Engineering, said the joint research also covers agentic AI workflows and AI-assisted software engineering. He identified model optimisation, multimodal AI and knowledge retrieval among the areas on which the institute and IBM are working.
IISc plans Granite models for energy analytics
IBM’s expanded programme with IISc, which began collaborating with the company in 2021, includes agentic systems and AI for scientific and industrial applications. The partners plan to study workflows that coordinate AI applications across hybrid-cloud environments while managing performance, cost and operational complexity.
For energy analytics, the researchers plan lightweight time-series foundation models based on IBM’s Granite model family. The proposed applications include forecasting, anomaly detection, load disaggregation and energy optimisation. IBM and IISc also plan to contribute benchmark datasets for research in the field.
The energy work gives the IISc programme a defined industrial application for Granite. Forecasting and anomaly detection concern changes in data over time, while load disaggregation and optimisation address energy use. IBM and IISc plan to develop the energy models and contribute benchmark datasets for wider research.
Yogesh Simmhan, a professor in IISc’s Department of Computational and Data Sciences, said the partners would explore cloud-native platforms spanning quantum-classical computing, edge AI and agentic workflows. The institute’s programme therefore includes both specialised models and systems intended to coordinate different computing environments.
IBM and IISc explore quantum-HPC algorithms
Quantum computing forms a further strand of the IISc collaboration. IBM and the institute plan to investigate how quantum and high-performance computing can be orchestrated in scientific workflows. The algorithm research will examine whether approximation-tolerant classical diagonalisation algorithms can improve quantum subspace iteration methods.
Amith Singhee, director of IBM Research India and chief technology officer of IBM India and South Asia, said the collaborations combine research at the two institutions with IBM’s scientific work. IBM, IIT Bombay and IISc have previously worked together across artificial intelligence, hybrid cloud, distributed systems and quantum computing.
Under the expanded IISc programme, researchers will also examine quantum-centric supercomputing workflows alongside the proposed quantum-HPC algorithms. The work will use approximation-tolerant classical diagonalisation algorithms in an effort to improve quantum subspace iteration.
