howdy, i'm dhiyaan

Led the design and development of a benchmark to evaluate how large language models handle demographic variation in medical question-answering tasks.

• Authored DiversityMedQA in collaboration with UC Berkeley EECS grads, introducing a new benchmark for evaluating demographic bias in medical QA systems.
• Accepted to NLP for Positive Impact (NLP4PI) Workshop at EMNLP 2024; presented at EMNLP and published in the ACL Anthology.
• Also accepted to the peer-reviewed Advancements in Medical Foundation Models (AIM-FM) Workshop at NeurIPS 2024 and the Southern California NLP Symposium 2025.
• Cited by NIH researcher in Ensuring Safety and Trust: Analyzing the Risks of Large Language Models in Medicine.

Worked under UT Austin ECE Professor Dr. Sriram Vishwanath and gained exposure to startup funding, product validation, and team-building.

• CSL is a 6-week Web3/Blockchain accelerator sponsored by Lightspeed Venture Partners and Polychain Capital, organized by Chainhub Foundation.
• Managed logistics for a Web3 accelerator backed by Google Web3, enabling collaboration between 14 startups, professors, and VCs.
• Developed and maintained the program's website used by investors for evaluating participating startups.
• Coordinated speeches and events featuring leading voices in blockchain and AI industries.

Akash Network is a decentralized cloud computing platform enabling low-cost, secure compute marketplaces.

• Developed an embedded system to verify GPU resource allocation for the Akash blockchain, ensuring compute reliability and resilience.
• Collaborated with UT Austin ECE PhD students and studied secure enclave architectures, including TEEs, RISC, and CISC systems.
• Built randomized protocols to verify GPU participation and defend against Sybil attacks.