OPEN TO WORK ✦
ML ENGINEER INDIA / WORLDWIDE DATA SCIENTIST STUDENT @ LPU

Models built
from scratch.

Krish Kumar — building intelligent systems from the ground up.
Deep learning, data pipelines, and the occasional Python game.
Yes, I mean scratch. No torch.autograd shortcuts here.

B.Tech CSE (AI/ML) @ Lovely Professional University
ABOUT ME
∞
Things being cooked simultaneously
🍳 REAL COOKING 💰 MONEY-MINDED 🚶 WALKING 🎮 GAMING ⚡ FAST LEARNER 🧠 FROM SCRATCH

I'm Krish Kumar — a student at Lovely Professional University pursuing B.Tech in CSE with a specialisation in AI & ML. I'm learning deep learning and data structures & algorithms while genuinely keeping joy in the process.

I'm built different: money-motivated, curiosity-driven, and always cooking something — literally and figuratively. Whether it's a neural network framework, a data pipeline, or an actual dish, I believe in building things from scratch before reaching for the wrapper.

When I'm not staring at loss curves, you'll find me on a walk, in the kitchen, or dreaming up the next thing that might — hopefully — make me some money.

WHAT I DO

I cook data
for a living.

01
ML Models
Building machine learning models from scratch or with frameworks. Neural nets, regression, classification — I understand every weight, gradient, and activation inside the black box.
02
Data Pipelines
Designing efficient data management frameworks that simplify the path from raw data to insight. Clean, composable, readable pipelines that actually make sense three months later.
03
Python Dev
From custom ML frameworks to Python games — building things that work beautifully. Strong OOP, clean architecture, and a genuine love for code that reads like a recipe.
SELECTED WORK

Things I've built.

01
Mini Neural Network Framework
A neural network engine built entirely from scratch. Supports forward pass, loss calculation, backpropagation, optimisation, and sequential model composition — no PyTorch, just pure first-principles understanding.
Python NumPy OOP
↳ github
→
02
DataManager Framework
A framework that makes data-related jobs simple and accessible for everyone. Clean API for building full data pipelines without the usual boilerplate — because life's too short for messy data wrangling.
Python Pandas NumPy OOP
↳ github
→
03
Linear Regression from Scratch
A clean NumPy implementation of linear regression with gradient descent. Zero sklearn. Full understanding of the mathematics — because using a black box you don't understand is just cheating yourself.
Python NumPy
↳ github
→
04
Multi-class Logistic Regression
A multi-class classifier built from first principles. Softmax, cross-entropy loss, and gradient descent — all hand-rolled with NumPy. Handles real multi-class problems, no training wheels.
Python NumPy Softmax
↳ github
→
05
Elastic Net Regression
Full elastic net with multiple optimisers (Adam, Lion, SGD) and learning rate decay strategies (Cosine, Step, and more). A proper ML toolkit built entirely from scratch — with real production-grade flexibility.
Python NumPy Adam Lion SGD Cosine Decay Step Decay
↳ github
→
TECH STACK

Tools I trust.

Python NumPy PyTorch Pandas C / C++ SQL NoSQL FastAPI Django Deep Learning Data Structures Algorithms OOP Git
JOURNAL

Thoughts
I've cooked.

COMING SOON
01
Deep Learning
Why I Built a Neural Net Before Touching PyTorch
What you actually learn when you implement backprop from first principles — and why it permanently changes how you use frameworks.
COMING SOON
02
Data Engineering
Data Pipelines Don't Have to Be Ugly
A guide to building clean, readable data pipelines that your future self will actually thank you for — no spaghetti allowed.
COMING SOON
03
Life & Learning
Learning ML Like Cooking: From Scratch Always Wins
On the philosophy of understanding fundamentals before using abstractions, and why the harder path up front pays dividends forever.
CONTACT
Let's cook
something real.

Open to internships, research collaborations, freelance ML work, and genuinely interesting problems. Especially the kind that pay. Drop a line — I reply fast.

GH
GitHub
→
LI
LinkedIn
→
CR
Credly
→
LC
LeetCode
→
@
Email
→