DEV Community

Sourav Sarangi
Sourav Sarangi

Posted on

My Journey Through the Kaggle Google 5-Day Intensive ML Sprint

This is a submission for the Google AI Agents Writing Challenge: [Learning Reflections OR Capstone Showcase]

My Learning Journey / Project Overview

Key Concepts / Technical Deep Dive

Reflections & Takeaways

Over the past week, I completed the Kaggle × Google 5-Day Intensive Program — a fast-paced, hands-on sprint that helped me dive into Python for Data Science, Machine Learning basics, and Kaggle-style workflows. Below, I’m sharing the full structure of the course, how I experienced each day, what I built, and the skills I gained. If you’re starting out in ML or thinking of trying Kaggle, this might help you decide if this path is for you.


📅 Course Structure & My Daily Experience

Day 1 — Getting Started: Python Basics + Kaggle Environment

✔️ Introduction to the Kaggle environment: Notebooks, datasets, competitions.

✔️ Brushed up on Python essentials — lists, dictionaries, loops, conditionals, functions.

✔️ First hands-on task: loaded a dataset using Pandas and performed basic exploration (head, shape, info).

My takeaway: Kaggle Notebooks are beginner-friendly, and running code live makes experimentation very straightforward.


Day 2 — Data Cleaning & Exploratory Data Analysis (EDA)

✔️ Learned data cleaning: handling missing values, removing duplicates, filtering outliers.

✔️ Explored data using Pandas: .describe(), grouping, filtering, summary statistics.

✔️ Performed preliminary visualization to observe data distributions and relationships.

My takeaway: Investing time in clean, well-explored data is critical — it lays the foundation for good ML results.


Day 3 — First Machine Learning Models (Baseline)

✔️ Understood the ML workflow: splitting data into training and test sets, fitting models, evaluating performance.

✔️ Built baseline models using Scikit-Learn:

Linear Regression (for regression tasks)

Decision Trees

Random Forests

✔️ Ran a quick mini-competition/prediction task on a real dataset.

My takeaway: Even baseline models — with minimal tuning — can deliver surprisingly decent results on real-world data.


Day 4 — Enhancing Models: Feature Engineering & Hyperparameter Tuning

✔️ Practiced feature engineering: generating new features, encoding categorical variables, scaling when required.

✔️ Applied hyperparameter tuning and cross-validation strategies to improve model performance.

✔️ Learned about the importance of model interpretation and avoiding overfitting.

My takeaway: Often, smarter features and better validation improve performance more than choosing a more complex model.


Day 5 — Final Project: End-to-End Pipeline + Submission

✔️ Built a complete ML pipeline: Data loading → cleaning → exploration → feature engineering → model training → evaluation → prediction.

✔️ Generated submission.csv and submitted to a real competition on Kaggle.

✔️ Witnessed the model’s score and placement on the leaderboard — first “real” ML submission.

My takeaway: Going from zero to a full submission in 5 days is possible — and hugely motivating. It turns theory into a tangible outcome.

Top comments (0)