About

The engineer behind the stack

Veda samhith

AI Engineering & Research

B.Tech in Aerospace Engineering
IIT Kharagpur

I am an Aerospace Engineering undergraduate at IIT Kharagpur transitioning into AI Engineering and Research. I believe the best way to understand complex systems is to build them from first principles. Over the past year, I have systematically reconstructed the entire modern LLM stack through 34 engineering projects— from tokenization to complete end-to-end systems. This journey has given me deep, end-to-end understanding of how language models work, where they break, and how to optimize them. I am particularly interested in efficient inference, long-context systems, and making AI more interpretable and safe.

Technical Stack

PyTorch
CUDA
Transformers
Deep Learning
Distributed Training
Model Optimization
Python
C++
Systems Programming
MLOps
Research Engineering
Technical Writing

Engineering Philosophy

1

Build to Understand

I believe the deepest understanding comes from implementation. Every project in this portfolio was built from scratch, not assembled from existing libraries.

2

Systems Thinking

LLMs are not isolated algorithms—they are complex systems. Understanding the interactions between data, training, inference, and serving is essential.

3

Research Mindset

Every implementation is an experiment. I document hypotheses, measure results, and iterate based on evidence rather than intuition.

4

End-to-End Ownership

From data curation to model serving, I take ownership of the entire pipeline. This ensures deep understanding of every component and their interactions.

Connect