Topic
Generative & Agentic AI
Exploring efficient LLM fine-tuning, mixture-of-experts architectures, and systems where language models interact with external tools and applications.
Generative & Agentic AI represents the transition from understanding how language models are built to exploring how they can be adapted and connected to real-world systems.
My work in this area spans three related directions.
First, I explored parameter-efficient fine-tuning, using QLoRA to adapt an instruction-tuned Llama model without requiring full-model training.
Second, I explored Mixture-of-Experts architectures, implementing ideas inspired by DeepSeek-V3 within SmolLM2 to understand how conditional computation can increase model capacity.
Finally, I explored agentic interaction, connecting an LLM with external tools to control LibreOffice Draw through an MCP-based workflow.
Together, these projects helped me think about modern AI systems as more than isolated models:
Foundation Model → Adaptation → Reasoning → Tool Use → Environment Interaction