> Initializing profile.md ...
> Loading neural weights ██████████ 100%
> Model ready. Hello, World.class MLEngineer:
def __init__(self):
self.name = "Krish Kumar"
self.role = "Machine Learning Engineer"
self.location = "Lucknow, Uttar Pradesh, India, Earth, Milky Way 🌍"
self.languages = ["Python", "C", "SQL", "Bash"]
self.focus = ["LLMs", "Computer Vision", "MLOps", "RL"]
self.currently = "Building my own mini neural network framework"
self.open_to = "Collaborations, Research, Freelance"
def __repr__(self):
return "grad_descent(curiosity=∞, lr=0.001)"
- Python (ML systems, NumPy-based implementations)
- C (low-level understanding, memory & performance basics)
|
Minimal deep learning framework implementing automatic differentiation, backpropagation, and tensor-based computation, with support for configurable layers, loss functions, and optimizers. |
A fully customizable Logistic Regression implementation built from scratch using NumPy.
Supports binary + multiclass classification, multiple optimizers, regularization methods, and preprocessing utilities. |
|
End-to-end Elastic Net regression implemented from scratch in NumPy with mini-batch gradient descent and model evaluation. |
Multivariate linear regression implemented entirely from scratch using vectorized gradient descent in NumPy. Includes manual cost computation, gradient derivation, feature scaling, and convergence analysis. |
[2025-Q2] 🔬 Mini Neural Network Framework from scratch
[2025-Q1] 🧞♂️ Logistic regression from scratch vs sklearn (~accuracy)
[2024-Q4] 🏆 2482th place — Kaggle House Prices - Advanced Regression Techniques (scored 0.14453)
[2024-Q3] 📈 Elastic-Net Regression vs sklearn
[2024-Q2] 📈 Linear Regression vs sklearn
┌─────────────────────────────────────────────────────────────┐
│ │
│ ◈ Large Language Models & Efficient Fine-Tuning │
│ ◈ Multimodal Learning (Vision + Language) │
│ ◈ Reinforcement Learning from Human Feedback (RLHF) │
│ ◈ Neural Architecture Search (NAS) │
│ ◈ Edge AI & Model Compression / Quantization │
│ ◈ Interpretability & Explainable AI (XAI) │
│ │
└─────────────────────────────────────────────────────────────┘
while alive:
eat()
sleep()
train_models()
push_to_github()
repeat()"All models are wrong, but some are useful." — George Box