Syed's Blog

114 posts · blog home · terminal version · RSS

Pandas Tricks That Make You Look Like a Wizard

· data-science

Method chaining, query, assign, groupby with named aggregations and other small habits that turn messy notebooks into readable analysis.

Data Cleaning: The 80% of the Job Nobody Puts on LinkedIn

· data-science

Duplicates, missing values, five spellings of the same city and dates in three formats. The unglamorous work that decides whether your analysis is right.

The Mean Lied to You: Meet the Median

· data-science

When a billionaire walks into a café, the average income in the room becomes enormous and completely useless. Why skewed data needs the median.

Correlation Is Not Causation (Ice Cream Didn't Cause That Shark Attack)

· data-science

Ice cream sales and shark attacks rise together. Banning ice cream will not save swimmers. A friendly guide to confounders and how not to embarrass yourself in a meeting.

Model Deployment: From My Laptop to Production Without Crying

· machine-learning

A trained model in a notebook is a science project. A model behind an API, with versioning and monitoring, is a product. The steps in between.

Cross-Validation: Trust Issues, But Make It Statistics

· machine-learning

One train/test split can be lucky. Cross-validation asks the same question five times so you can trust the answer, with a few traps for time series and grouped data.

Quantum Machine Learning: Hype, Hope and a Lot of Qubits

· machine-learning

What quantum computers might genuinely do for machine learning, what they can't do yet, and how to tell research from marketing.

Hyperparameter Tuning, or: Expensive Guessing

· machine-learning

Grid search, random search and smarter search, plus the uncomfortable truth that tuning usually matters less than data.

Class Imbalance: When 99% Accuracy Means Nothing

· machine-learning

A fraud model that never predicts fraud can be 99% accurate. Here's why, and which metrics actually tell you the truth.

Vision Transformers for Satellite Images, Without the Jargon

· machine-learning

How a model that reads images in patches can map vegetation, water and land use from space, and why satellite data is weirder than your phone photos.

Federated Learning: Training Models Without Peeking at Your Diary

· machine-learning

How phones, hospitals and IoT devices can train a shared model while the raw data never leaves home, and why it's harder than it sounds.

Feature Engineering Is Where the Magic (and the Pain) Lives

· machine-learning

Fancy algorithms get the headlines, but the features you give them decide most of the result. Practical examples with dates, ratios and categories.