HYDERABAD: At a time when artificial intelligence (AI) reshapes sectors from healthcare and education to defence and finance, Prof Priyesh Shukla of IIIT Hyderabad is working on a hardware solution that aims to ease the strain on the machines powering these systems.
An assistant professor, Shukla is developing TVARAK-AI, a scalable and heterogeneous accelerator chip designed for sustainable inference in next-generation AI models. His work focuses on addressing the rising computational demands, memory constraints and energy use associated with modern AI systems, which are becoming larger and more complex.
Shukla research receives national recognition with the Prime Minister Early Career Research Grant (PMECRG) awarded by the Anusandhan National Research Foundation (ANRF). He is selected among 700 researchers from over 6,000 proposals across the country. The project sits at the intersection of artificial intelligence, semiconductor design and efforts towards technological self-reliance.
Named after the Sanskrit word ‘Tvarak’, meaning to accelerate, the chip is patented and designed to provide an alternative to conventional Graphics Processing Units (GPUs), which are often expensive, energy-intensive and largely controlled by global technology firms. The proposed system seeks to improve efficiency by managing computing loads and memory access more effectively, while reducing overall power consumption.
“The project goes beyond designing a chip and aims to build an integrated AI ecosystem combining hardware architecture, software tool chains, machine learning compilers and deployment frameworks,” Shukla tells TNIE.
Currently in the prototyping phase, the research builds on his earlier work in in-memory computing-based accelerators for edge AI. This approach allows processing closer to where data is generated, which reduces latency and energy usage while supporting deployment in resource-constrained settings.
The proposed system is expected to support applications across rural healthcare, autonomous drones, mobile devices and low-connectivity environments. By reducing reliance on imported AI hardware, it also aligns with India’s broader push for domestic capability in semiconductor and AI infrastructure.
Amid rising investment in semiconductors and AI, TVARAK-AI could shift India from a consumer to a developer of core technologies, if commercialised.