Experience

Agentic AI Systems

• Contributed to the development of a three-stage agentic AI workflow: compile, implement, and execute, supporting structured workflow automation.
• Assisted in building and testing modules that transform high-level workflow plans into atomic functions and execute them with LLM-powered agents.

Weight Generation for Large Language Models

• Conducted a comprehensive literature survey on generative models for weight generation and alternative approaches to weight-space learning in neural networks.
• Ran and analyzed experimental results from existing codebases, assisted in debugging and reproducing key experiments to validate methodologies.
• Investigated weight distribution properties (kurtosis, compressibility) and contributed to the design and implementation of latent-space fusion experiments.
• Co-authored a paper (ICLR 2026): LS-Merge: Merging Language Models in Latent Space, proposing a framework for merging heterogeneous large language models in latent space.

LLM Agent on Hyperparameter Optimization

• Conducted a literature review on state-of-the-art hyperparameter optimization techniques for large language models (LLMs).
• Analyzed existing codebases and replicated experiments to understand optimization workflows.
• Implemented and tested existing optimization methods to assess their impact on model performance.

Hyperparameter Optimization

• Developed and implemented baseline optimization algorithms, including BOHB, DEHB, and FSBO to efficiently optimize complex black-box functions.
• Evaluated the performance of these algorithms on benchmark problems and real-world applications.
• Co-authored a paper (NeurIPS 2025): Cost-Sensitive Freeze-Thaw Bayesian Optimization for Efficient Hyperparameter Tuning.

AI Programming with Python

• Participated in the AWS DeepRacer Student League and received the AWS AI & ML Scholarship.
• Completed a collaborative virtual course on programming tools and techniques fundamental to machine learning.