About me

Hello! I am Ankit Pratap Singh. I am a PhD candidate in the Department of Electrical and Computer Engineering at Iowa State University, advised by Professor Namrata Vaswani. I am currently working as an AI Researcher at Thoughtworks, focusing on developing intelligent and scalable data generation systems. My research interests include Optimization, Agentic AI, Reinforcement Learning, Multi-Task Representation Learning, and Secure Federated Learning. During my Ph.D., I worked on Secure Federated Learning, Reinforcement Learning, and Multi-Task Representation Learning, leading to publications in ICML and IEEE Transactions on Information Theory (T-IT). At the IFC (World Bank Group), I developed a multi-agent LLM system for the Financial Spreading problem, integrating document understanding and reasoning for large-scale financial analysis. I am continually seeking and open to collaborations on impactful projects at the intersection of AI, optimization, and distributed learning.

A Mathematical Perspective on Transformers: Sometimes, a mathematical framework makes it easier to understand concepts than relying on abstract explanations.

Recent News

• Joined Thoughtworks as AI Researcher
• Joined International Finance Corporation (World Bank) as ML Researcher
• Paper on Byzantine-Resilient Federated Alternating Gradient Descent and Minimization for Partly-Decoupled Low Rank Matrix Learning is accepted in ICML 2025
• Paper on Secure Algorithms for Vertically Federated Multi-Task Representation Learning is accepted in ISIT 2025
• Paper on Byzantine-resilient federated pca and low rank column-wise sensing is accepted in IEEE Transactions on Information Theory Journal
• Paper on Byzantine Resilient and Fast Federated Few-Shot Learning is accepted in ICML 2024
• Paper on Byzantine-Resilient Federated Principal Subspace Estimation is accepted in ISIT 2024

Research Interests

• Agentic AI
• Robust Machine Learning
• Federated learning
• Information Theory