Akash Kamalesh

I'm a B.Tech Computer Science and Engineering graduate from PES University, Bengaluru. Currently conducting independent research on building scalable and efficient AI systems, with particular emphasis on representation learning, multimodal alignment, contrastive learning, and fine-tuning strategies for foundation models.

I've worked on methods like UnoLoRA for efficient multitask adaptation, explored grounding information better in vision-language models, and developed cross-lingual sparse Mixture of Experts architectures. My broader goal is to make foundation models more aligned, interpretable, and adaptable across tasks and modalities.

In addition to research, I've applied these ideas in real-world settings through roles at IISc, Swiggy, Nokia, and Cisco, where I worked on applying generative AI across practical domains such as finance and healthcare, building conversational recommender systems for personalized item suggestions and solving graph network problems.