Ishaan Domkundwar
E1102, 24K Glitterati, Vishal Nagar, Pune 411027, India
Hello! I am an IB Diploma student at Mahindra International School in Pune, where I live with my parents, grandparents and my younger sister. Browse my website to learn more about me.
Resume
AI Research
Browse my research around AI agents safety and LLM performance benchmarks.
AI Agent Safety Research at SuperAGI
AI Agent Safety Research at SuperAGI During my research at SuperAGI, I focused on addressing the critical issue of AI safety. I investigated the limitations of current AI systems and developed three safety frameworks aimed at reducing risks associated with AI operations. These frameworks were designed to enhance the reliability of AI agents across a wide range of applications. By testing these methods on actual AI systems, I was able to demonstrate how they can significantly improve safety protocols. My work culminated in the publication of a research paper on arXiv.org, where I outlined these frameworks and proposed solutions to mitigate the inherent risks, ultimately contributing to the advancement of safe AI deployment.
Read the full research paper here OR Download PDF here.
LLM Performance Benchmarking Research
LLM Performance Benchmarking Research As part of my research on large language models (LLMs), I worked under the guidance of a PhD student at MIT to improve the logical reasoning capabilities of these models. My primary focus was on identifying and addressing weaknesses in sequence prediction, which I evaluated by analyzing over 6,000 arithmetic sequences. Through this research, I explored innovative methods to enhance the predictive accuracy of LLMs, significantly advancing their performance in complex tasks. The findings from this research were published in a paper, where I presented a set of strategies for improving the logical reasoning abilities of LLMs, offering a valuable contribution to the field of AI research.