Prannoy Pilligundla

I work on the Ads Model Serving team, architecting systems that power ad ranking across Meta's products. Before this, I spent four years at AWS building cloud ML platforms like Panorama and Lookout for Vision, and started my career at NVIDIA working on CPU architecture.

I have a Master's in Computer Science from UC San Diego, where I researched neural network quantization and compiler optimization under Dr. Hadi Esmaeilzadeh. My work has been published at ICML, ICLR, and NeurIPS.

I did my undergrad at BITS Pilani in India, where I was a two-time Google Summer of Code participant—porting jor1k to support RISC-V architecture with LowRISC, and building a LyX to Word converter.

Work

Research

Graduate research at UC San Diego focused on making ML models faster and more efficient.

• Divide and Conquer: Leveraging Intermediate Feature Representations for Quantized Training of Neural Networks ICML 2020
• Chameleon: Adaptive Code Optimization for Expedited Deep Neural Network Compilation ICLR 2020
• ReLeQ: A Reinforcement Learning Approach for Deep Quantization of Neural Networks NeurIPS 2018 Workshop Spotlight

Connect

Find me on GitHub, LinkedIn, Google Scholar, or email me at prannoy.bits@gmail.com.