Kartik Hegde

- Leading the ChipStack integration with Cadence's portfolio of products - Significantly scaling the distribution & adoption of ChipStack's technology - Influence and shape the Agentic AI efforts at Cadence

ChipStack is leading the generative AI-driven revolution in chip design. - Built a team of 20+ amazing engineers across chip design, AI, and software. Can't be prouder of the team we have brought together! - Raised over $7m in seed from top VCs like Khosla Ventures - 10+ enterprise customer deployments, closed deals worth $4m+, live on flagship projects - Product focused on verification: AI-driven formal, unit testing, and UVM - Led the M&A process and delivered strong ROI to investors and everyone involved Read more at www.chipstack.ai

Majority of my Ph.D. work was done in collaboration with exceptional colleagues at Nvidia research. My work broadly covered the following: - Measuring and improving the performance of Nvidia GPUs for sparse tensor algebra - Proposing and prototyping alternate architectures for select deep learning workloads Paper: https://www.kartikhegde.net/publication/extensor/

- Exploring new architectures for irregular workloads such as recommendation systems - Implementing prototypes with simulators and RTL to estimate PPA

- Applying deep learning for compilers for power and performance-efficient mapping of algorithms - Effective use of gradient-descent for black-box optimization problems - Led to over 40% improvements in EDP of target accelerators Paper: https://www.kartikhegde.net/publication/mindmappings/

Year 1: - System-level verification of Arm SoCs, targeting server class systems - Coherency verification for CPU-GPU coherent interconnects - IP-level verification of Arm's mobile-class GPUs and other media IP (video, image, processors) - Using ML for finding patterns of RTL bugs Year 2: - Architect and implement a prototype low-power ML accelerator at Arm research, targeting deep learning - Led to a <10mw design that heavily influenced Arm's future trajectory in building NPUs

Supercomputer Education Research Center - Exploring ARM processors for High performance Computing.

Systems and software group, Processor division

I am exploring how computer architecture can continue the growth it has been achieving beyond the end of Moore's law. In particular, I envision the future computing to be accelerated by domain specific massively heterogeneous architectures. Currently, I am developing efficient accelerators for deep learning, mainly focused on videos.

Excellent academic Record and strong Laboratory experience, part of various research projects.