AI / GenAI Engineer in the Making
Chennai, India
LinkedIn | Email | Medium | Hugging Face
I build AI systems by understanding the mechanics underneath them, not just calling APIs. Right now I am focused on neural networks, LLM applications, RAG systems, and agent workflows.
A character-level language model and Adam optimizer built from scratch in pure Python — no PyTorch, no NumPy — trained on the Bhagavad Gita's Krishna teachings.
- my own autograd engine powering a recurrent char-level model
- Adam implemented by hand, on top of hand-written backprop
evidence/folder with training logs, generated samples, and a loss plot
Why this project matters:
- shows implementation-level understanding of backprop, optimization, and sequence modeling
- zero dependencies means the mechanics are fully visible and inspectable
- demonstrates that I can take a from-scratch build from idea to documented artifact
- autograd-charlm-scratch: char-level LM + Adam from scratch in pure Python, trained on Krishna teachings
- micrograd-from-scratch: autograd engine, neural network stack, optimizers, schedulers, and benchmark artifacts
- VENA: vision-enabled navigation agent
- ted-talk-recommender: content-based recommendation system
- Merged OSS contribution: langchain-google-genai integration tests in RobotecAI/rai (PR #740, March 2026)
- I rebuilt micrograd from scratch because backprop was annoying me
- I Built a Character-Level Language Model from Scratch. It Mostly Learned to Overfit.
- LLM applications
- RAG systems
- AI agents
- deep learning fundamentals
- open source contributions (1 merged PR, 2 in review)
I am actively looking for internships and early-career opportunities in:
- Generative AI
- LLM engineering
- AI agent engineering
- machine learning engineering
If you are hiring for internships, research work, or project-based collaboration, reach out: rangaprasath22@gmail.com
