#machine-learning

Are Tabular Foundation Models Ready to Replace Gradient Boosting Models?

A practical comparison covering architecture, TabArena benchmarks, latency, licensing, and runnable code.

I built an AI to beat the solver at airline crew recovery. It lost, and that loss became the product

Re-crewing a storm-grounded hub is a 4-to-12-hour scramble under FAA Part 117 and union rules. The fix was never a smarter optimizer. It was

Spam Email Detection using LSTM

In this project, you will build a deep learning model to classify messages from SMS or email as spam or ham (not spam), using a Long…

Uncertainty Quantification in AI (Part — 2)

Chapter 2 — The Idea Behind Conformal Prediction

Uncertainty Quantification in AI (Part — 1)

The following is series of posts on Uncertainity quantification in AI systems and discussion on some of the statistical methods to overcome…

Three Half-Truths About Getting Your Data AI-Ready

Consolidate, clean, lock down. Three real projects that each leave the thing blocking production exactly where it was.

Adaboost Algorithm

“One weak model makes mistakes → the next model is forced to focus more on those mistakes.”

Training Models Bigger Than Your GPU: A Complete Guide to DeepSpeed

How Microsoft’s Optimization Library Turned Extreme Scale Deep Learning Into a Practical, Everyday Tool

Mask R-CNN in Deep Learning: A Complete Guide to Object Detection and Instance Segmentation with a…

Introduction

Generating Realistic Mars Images Using DCGAN: A Deep Learning Guide

Learn how Deep Convolutional Generative Adversarial Networks (DCGANs) can generate realistic images of Mars by learning directly from real…

AI Isn’t Magic. It’s Hungry

AI Isn’t Magic. It’s Hungry.

Exploring AI Hand Tracking with MediaPipe: My First Hands-on Experience

Exploring AI Hand Tracking with MediaPipe: My First Hands-on Experience

Building an End-to-End Fraud Detection ML Pipeline on AWS

Introduction

Mac Mini M4 vs RTX 5090 vs Cloud GPUs for Local AI in 2026

The $1,799 box beats the $4,000 one for most local AI, and the reason is one number: 32.

AI Detectors Can Score 97% and Still Be Wrong

A 97% contest score can mean “aced one practice test” — not proof of authorship. False positives, new models, and humanizers punch holes in…

TD Error as Advantage

Distribution-Free Uncertainty

Neural networks are pathologically overconfident. A model trained to minimize cross-entropy will cheerfully predict a 99% probability for…

AI can’t save you from your own stupidity.

Reading the news is not about knowing more events. It is about noticing what the world is beginning to need. Somewhere inside that…

Bag of Words: The Unglamorous NLP Technique I Still Reach for First

Before neural machine translation, before transformers, before any of the fancy stuff — every NLP project I’ve built started with this one…

Beyond Building a Classifier: What I Actually Learned Building an ML Classification Project

This is not a tutorial. It is a documentation of what I built, what broke, what surprised me, and what I would do differently, written…

How Does Gmail Know It’s Spam Before You Do?

Every morning, I open Gmail expecting to see the emails that matter.

MoE Cost Analyzer: Benchmark Dense vs Mixture-of-Experts LLMs on Your Own Prompts with Real API…

Every team running LLMs in production eventually asks the same question: should we switch to a Mixture-of-Experts model? The MoE…

AI, Machine Learning ও Deep Learning এর মধ্যে পার্থক্য কী?

“AI আর Machine Learning কি একই জিনিস? Deep Learning আবার কী?”

Porting a PyTorch CNN to Rust

Youtube video: https://youtu.be/JxXepbop52A?si=YAKBSOyMfB83C-l9

You Don’t Need Data from Everyone to Make Smart Decisions.

Imagine you’re asked to predict the winner of a national election.

Data Types in Python

Essential Python datatype every beginner need to know

The shift(-1) Bug: How a One‑Character Mistake Turned a Random Model into a 77,000% Backtest

A reproducible case study on why naive machine learning almost always fails on financial data, and what a proper walk-forward validation…

DAY 2: DECODING AI — मशीनों के सोचने का रहस्य (मशीनें कैसे सीखती हैं?)

#DailyLearnAIwithSarita | सफर का दूसरा दिन

Why Old AI Models Suddenly Feel Stupid.

The strange psychology of moving baselines, model drift, and our shrinking patience for yesterday’s magic.

SC20: Numerical Integration

Find the whole linear algebra series here

Temporal Difference (TD) Learning is Beautiful

To master reinforcement learning (RL), you have to transition from planning with a known model (like Dynamic Programming) to learning…

Multi-Agent Reinforcement Learning for Smart Cities and Beyond

How Intelligent Agents Are Learning to Coordinate Traffic, Energy, Robots, and Entire Urban Ecosystems

The Hidden Compliance Gap Between What AI Governance Says and What ML Infrastructure Does

Your models are good. Can your systems account for what they did? An infrastructure engineer’s reading of the EU AI Act’s technical…

I Compared Toolyzo and Originality.ai Using the Same 5 Writing Samples

AI detectors are becoming part of everyday workflows for publishers, students, marketers, and businesses.

5,811 Dermatology Cases Expose the Clinical AI Demo Trap

Google warns MedGemma needs use-case validation; a May 2026 Yale cohort shows why real consults, context, and handoffs must decide the…

Lets Investigate The Hype Around Facebook’s Big Comeback With Muse Spark 1.1, Fact or Fiction

Meta put Muse Spark 1.1 into public preview on July 9. The part I care about is not the usual claim that a new model can plan, code and use…

Kimi K3 Released : Beats Claude Fable, GLM 5.2, GPT 5.6-Sol

How to use Kimi K3 for free?

Gaussian Mixture Models (GMM) Explained: A Simple Guide

Learn how Gaussian Mixture Models work in machine learning, how they use soft clustering, and why they go beyond simple normal…

How to Build a Software Engineering Portfolio

A software engineering portfolio is one of those things where the conventional advice (“just build projects and put them on GitHub”) is…

What Is Inference Engineering? The Layer Doing 80% of Your LLM Bill.

FP8 KV cache, prompt caching, quantization, speculative decoding, MoE routing. The five 2026 levers between your model and your invoice…

The Real Cost of AI Math Errors: What Happens If You Just Leave It

According to McKinsey, 42% of businesses using AI faced losses from bad outputs in 2025. AI math errors drive a large share of that damage.

From Numbers to Decisions: The Role of Statistics in Data

Table of Contents

How to use Kimi K3 for FREE ?

Kimi K3 : Biggest ever AI Model released

TinyML — Generative Adversarial Networks

From mathematical foundations to edge implementation

The Repository That Reviews Itself

Triage bots, disposable test boxes, pooled API budgets, and a review loop that calls itself, reconstructed from the source.

Where does Antigravity look for Hooks?

💡 This post is part of a series on Antigravity configuration:

How to implement GPT-2 from scratch like Karpathy

I spent two hours watching Andrej Karpathy build GPT from scratch, nodding along like it was clicking. Then I opened a blank file to build…

The Reason Mira Murati’s $2B Model Is Losing on Purpose

Everyone is reading Inkling’s benchmark table wrong. The one number that actually matters isn’t in it.

Enterprise RL: When Fast Vectorized Environments Beat Bigger Models

A dynamic-pricing intuition, tested with a controlled Tic-tac-toe benchmark

When creativity and art collide

When a machine can hand you a masterpiece in nine seconds, the thing you lose is the part you never noticed you had