Ashish Kumar

I wrote my first "Hello, World!" when I was 11, and that curiosity eventually grew into a deeper interest in ML, deep learning, and AI. Since then, I've explored LLMs, diffusion models, multimodal learning, NLP, CV, RL, world models, TTS, generative modeling, and agents, with a strong interest in understanding how different kinds of models learn, represent information, and generate outputs.

I especially enjoy making models from scratch. Whether it's an LLM, diffusion model, NLP system, TTS model, or another deep learning model, I like starting from the idea, implementing the architecture, preparing the data, training it, and seeing how far I can take it. I experiment with different architectures, objectives, datasets, and training setups, reproduce ideas from papers, and then modify them to see what else is possible. I also enjoy robotics, AR/VR, and embodied AI, especially where ML can turn these technologies into systems that can understand, generate, and interact with the world.

I believe AGI is possible, and a large part of why I study AI is to understand what it would take to get there. I want to work on the models, algorithms, and systems that move us toward more capable and general forms of intelligence, while exploring how those systems can eventually connect with robotics and spatial computing. For me, the goal is not simply to use existing AI, but to understand it deeply enough to build new models, test new ideas, and contribute to what comes next.

I also play Chess and do competitive programming, you can watch my profile here on CList.