I'm Haricharan, an AI systems engineer.
I like building things from scratch, whether that's training custom LLMs, building autonomous agent runtimes, or scaling distributed pipelines.
If it involves understanding how models actually work under the hood, I'm probably experimenting with it.
My main tech stack is Python, PyTorch, CUDA, Docker, and Linux. I spend most of my time around Transformers, distributed scaling (DDP, ZeRO), and agent frameworks like MCP.
I've got a few pinned projects below, but my favorite is my Autonomous Coding Agent Harness.
Facts:
- I prefer writing models from scratch over just calling APIs
- Hobbies: building side projects, exploring new AI research, gaming, anime