Kavana Venkatesh

👋 Hi there! Welcome to my corner of the internet! I am a PhD student at Virginia Tech. Previously, I earned my master’s degree in Data Science from Northeastern University, Boston. My research interests are agentic reasoning, safety, and interpretability in Large Language Models, with a focus on evaluation frameworks and grounded control across long-context, multimodal, and generative tasks. I also explore the intersection of NLP, vision, and retrieval to enable reliable and adaptive behavior in LLM-based systems. When I’m not busy exploring AI frontiers, you might find me geeking out over Large Language Models, enjoying fun books, or perfecting my latest cup of coffee.☕

A few fun things about me:

• 🌍 I love traveling and capturing stories through photography.
• 🐾 Obsessed with pets, especially dogs.
• 📚 Passionate about storytelling and making AI more accessible to everyone.

Feel free to explore my work, and reach out if you’d like to collaborate or just chat about AI and beyond! 😊

This summer, I enjoyed interning at Amazon's Air Science & Tech as an Applied Scientist Intern at their Bellevue, WA office, where I contributed to enhancing Amazon Air's daily demand planning using efficient RL and ML techniques!

🔬 Research

My research focuses on Large Language Models and extending the evolving technology in NLP to Computer Vision tasks, especially Diffusion Models.

I aim to develop interpretable and efficient generative models to enable safe consumption of AI at scale across diverse domains. Specific areas of focus include:

• Interpretability of Large Language Models: Understanding internal models and thought processes behind the reasoning of LLMs
• Agentic Framework and Tool Usage: Researching efficient agentic ecosystems for collaborative reasoning and smart tool usage in LLMs
• Evaluation and Benchmarking for LLMs: Developing innovative and dependable approaches to evaluate LLMs for complex scenarios
• Knowledge Distillation and Optimization of LLMs: Exploring novel optimization techniques to develop smaller and efficient LLMs
• Applications of LLMs in vision: Transferring the evolving techniques in NLP to vision, especially diffusion models for high-fidelity image generation

📰 News