Hello! I'm Prajakta Kini, a 2nd-year Master's student in Computer Science at the University of Colorado Boulder. My research interests lie in Natural Language Processing (NLP), with a focus on neuro-symbolic approaches, improving interpretability and alignment in large language models (LLMs), and meta-reasoning in LLMs.
Currently, I am advised by Dr. Maria Pacheco in the BLAST Lab, where I am exploring neuro-symbolic methods for meta-reasoning—developing ways for LLMs to evaluate and improve their own reasoning processes.
Additionally, I am advised by Dr. Alvaro Velasquez and Dr. Amrit Singh Bedi on AI misalignment and scheming detection, where I am developing benchmarks for evaluating covert misalignment in advanced AI systems.
From Sept 2024 – Aug 2025, I also worked with Dr. Theodora Chaspari in the HUBBS Lab on bias detection and fairness in mental health NLP. Our paper, "Assessing Algorithmic Bias in Language-Based Depression Detection: A Comparison of DNN and LLM Approaches", has been accepted at the BHI 2025 conference.
Before graduate school, I worked as a Software Developer at Uber and VMware in India for over four years, which gave me a strong foundation in large-scale software development. More recently, I interned as a Summer SDE Intern at PhonePe, USA (June–August 2025).
Outside of academics, I enjoy listening to music and experimenting with cooking, helping me unwind and explore creativity in different ways.
Recent News
- September 2025: Submitted Paper "Taming Misalignment in Generative AI Agents via Social Simulations: The Next Frontier for AI Safety" to ETA track, AAAI26 AAAI-26 Conference
- August 2025: Our Paper "Assessing Algorithmic Bias in Language-Based Depression Detection: A Comparison of DNN and LLM Approaches" accepted at IEEE-EMBS BHI 2025 Conference