DEV Community

Susheem Koul profile picture

Susheem Koul

Senior Software Engineer at Microsoft, based in Hyderabad. I like taking systems apart to see how they actually work. AI Architectures, Distributed Systems and some random thoughts.

Location Hyderabad, India Joined Joined on  Personal website https://susheem-k.github.io/ github website

Education

BITS Pilani

Pronouns

he/him

Work

Senior Software Engineer at Microsoft

Pin Pinned

How 8-bit quantization shrinks an LLM to half of its size — and why a single outlier weight can quietly ruin it
susheem-k profile

How 8-bit quantization shrinks an LLM to half of its size — and why a single outlier weight can quietly ruin it

#ai #llm #machinelearning #python
12
Comments 9
11 min read
I built an RPG that teaches Claude Code by making you actually use it

Graded by real-state checks instead of quizzes
susheem-k profile

I built an RPG that teaches Claude Code by making you actually use it

#ai #claude #opensource #node
18
Comments 10
3 min read
Training for quantization: QAT, 1-bit BitNet and 1.58-bit weights
susheem-k profile

Training for quantization: QAT, 1-bit BitNet and 1.58-bit weights

#machinelearning #llm #ai #deeplearning
7
Comments
8 min read

Want to connect with Susheem Koul?

Create an account to connect with Susheem Koul. You can also sign in below to proceed if you already have an account.

Create Account
Already have an account? Sign in
From Validating Code to Evaluating Behavior: Rethinking Testing for AI Agents
susheem-k profile

From Validating Code to Evaluating Behavior: Rethinking Testing for AI Agents

#ai #testing #agents #software
12
Comments 2
9 min read
How coding agents like Cursor quietly cut input costs by reusing KV states across turns — and what actually breaks the cache
susheem-k profile

How coding agents like Cursor quietly cut input costs by reusing KV states across turns — and what actually breaks the cache

#ai #webdev #llm
5
Comments 1
4 min read
loading...