About Me

Hey there, I'm Sai.

I'm a recent UC Berkeley graduate with a B.A. in Computer Science and Statistics, with a minor in Public Policy. I'm broadly interested in artificial intelligence and machine learning systems, especially at the intersection of alignment, reliability, and real-world deployment.

I'm currently building PlotViews, a company focused on using AI to automate and elevate commercial real estate visualization and marketing. In parallel, I continue to work on AI research, primarily around agentic systems, evaluation, and scalable AI frameworks.

My research interests include:

• AI alignment and safety
• Reliability and evaluation of LLM-based systems
• Agentic and compound AI systems

Previously, I co-developed Agent Arena, an interactive platform for evaluating LLM-based agents, and led development on Ember, a compositional framework for building Compound AI Systems and Networks of Networks. I am also a co-author of LLM Chess, a framework for evaluating LLM reasoning and planning through chess gameplay, which was published at the NeurIPS FoRLM Workshop.

I have prior industry experience as a software engineer at Arize AI, IBM (DataStax), and Priceline. I enjoy working at the boundary between research and production, especially on systems that need to be both theoretically grounded and practically useful.

Featured Research Projects

• LLM Chess: A framework for evaluating LLM reasoning and planning through chess gameplay, published at the NeurIPS FoRLM Workshop. Features a novel approach to testing language models' ability to understand game states, plan moves, and execute complex strategies in a zero-shot setting
• Agent Arena: An interactive platform for evaluating and comparing LLM-based agents. Agent Arena provides a systematic way to assess agent capabilities through battle-based comparisons
• Ember: A compositional framework for constructing Compound AI Systems and Networks of Networks (NONs), focusing on scalable and modular AI workflows
• Emotion-tune: A project focused on improving RLHF with emotion-based feedback. Built on EMILI (Emotionally Intelligent Listener), it uses real-time facial emotion recognition to enhance AI interactions and advance long-term AI safety through emotion-aware systems

Other Interests

• Sports: Golf, Basketball, Tennis, Pickleball, Snowboarding
• Games: Poker, Video Games, Anime
• Music: R&B, House, Hip-Hop
• NBA: Boston Celtics