Fahad Hashmi

📊 Exploratory Data Analysis of Student Academic Performance Dataset (Math, Reading, Writing)
Analyzed how gender, parental education, lunch type, and test preparation impact academic outcomes.
Uncovered insights on average scores, correlations, and performance gaps using real-world educational data.
Visualized results using Pandas, Seaborn, and Matplotlib in Jupyter Notebook.
📩 DM for access to the Jupyter Notebook or code walkthrough.

📊 Complete Exploratory Data Analysis of Students’ Academic Performance Dataset | Gender Gaps, Test Prep Impact & Socioeconomic Insights

As part of my data science portfolio, I conducted a comprehensive Exploratory Data Analysis (EDA) on a student academic performance dataset, which includes test scores in math, reading, and writing, along with key demographic features such as gender, parental education, race/ethnicity, lunch type, and test preparation status.

This end-to-end analysis involved cleaning and transforming real-world data, identifying score trends, and visualizing how socioeconomic and educational factors influence academic outcomes.

🔍 Scope of the Analysis:

🧹 Data Cleaning & Feature Engineering

Handled column renaming, formatting issues, and ensured correct data types.

Created a new total score column by combining math, reading, and writing scores.

Verified the distribution of scores and filtered any data inconsistencies.

👩‍🏫 Performance by Gender

Compared subject-wise averages for male vs. female students.

Visualized gender-based trends using boxplots and bar charts.

🧪 Impact of Test Preparation Course

Analyzed how completing a test prep course affected scores.

Found significant improvement in all three subjects for students who completed the course.

🎓 Effect of Parental Education

Investigated how a parent’s education level relates to student performance.

Identified a positive correlation between higher parental education and higher scores.

🍱 Socioeconomic Analysis via Lunch Type

Compared performance of students with standard lunch vs. free/reduced lunch.

Highlighted disparities suggesting socioeconomic impact on learning outcomes.

🌍 Racial/Ethnic Group Trends

Analyzed average scores for each racial/ethnic group (Groups A to E).

Explored patterns of academic strength and disparity among different groups.

📈 Score Distribution & Correlations

Visualized distribution of each subject’s scores using histograms and boxplots.

Used heatmaps show strong correlation between reading and writing scores.

🧠 Tools & Skills Demonstrated:

Data manipulation with Pandas

Advanced visualizations with Seaborn & Matplotlib

Statistical analysis of categorical and numerical variables

Insight generation for educational data-driven decision making

📌 Full code, visualizations, and Jupyter Notebook will be available on GitHub.

🔗 https://github.com/ifahadhashmi/Pandas-EDA-Mastery

🔗 Feel free to connect or message me for collaboration or feedback!

📊 Complete Exploratory Data Analysis of Pakistan’s Population Dataset (1998–2017) | Urban-Rural Dynamics, Growth Trends & Demographic Insights
As part of my data science portfolio, I conducted a comprehensive Exploratory Data Analysis (EDA) on Pakistan's official population dataset, covering all administrative levels — from provinces to tehsils. This end-to-end analysis involved cleaning complex census data, transforming raw variables, and generating insightful visualizations to uncover key demographic patterns and regional growth trends.

🔍 Scope of the Analysis:
🧹 Data Cleaning & Preprocessing

Handled missing values, standardized inconsistent formats, and structured hierarchical region mappings.

Converted string-based numerical data (e.g., population, area, household size) into usable formats.

📍 Provincial & District Demographics

Identified the most populous provinces and districts.

Analyzed area coverage and population density at each level using comparative bar charts and choropleth maps.

🌆 Urban vs Rural Dynamics

Evaluated urban-to-rural population ratios across provinces.

Compared average household sizes between rural and urban regions.

Highlighted areas with the highest urban or rural population growth.

📈 Temporal Growth Analysis (1998–2017)

Assessed district- and division-level growth rates across the census years.

Investigated which divisions are expanding fastest and where urbanization is accelerating.

⚖️ Gender Ratio & Correlation Insights

Visualized gender balance across regions.

Explored correlations between population growth, gender ratios, area size, and household structure using heatmaps and pairplots.

📊 Comprehensive Regional Comparisons

Compared divisions within provinces to uncover internal growth disparities.

Quantified urban vs rural population proportions for each province.

🧠 Tools & Skills Demonstrated:
Data wrangling with Pandas

Advanced visualization using Seaborn and Matplotlib

Analytical storytelling and statistical insight extraction

Practical handling of real-world, multi-layered demographic data

📌 Full code, visualizations, and Jupyter Notebook will be available on GitHub.
[https://github.com/ifahadhashmi/Pandas-EDA-Mastery]

🔗 Feel free to connect or message me for collaboration or feedback!

📊 Complete Exploratory Data Analysis of Google Play Store Dataset | A Comprehensive Breakdown

As part of my data science practice series, I performed a complete EDA on the Google Play Store dataset to uncover key insights related to app categories, user behavior, and monetization strategies. This project involved cleaning complex real-world data and extracting actionable patterns to support decision-making in mobile app analytics.

Key Highlights of the Analysis:

🔍 Data Cleaning & Preprocessing

Handled missing values and inconsistent formats (e.g., size units like ‘M’ and ‘k’, and price symbols).

Converted categorical and string-based numerical fields to appropriate data types for analysis.

📊 App Category Distribution & Popularity

Analyzed the spread of apps across categories to identify dominant sectors in the Play Store.

Used bar plots and count plots for clear visual interpretation.

🌟 User Engagement Patterns

Explored relationships between ratings, reviews, and installs.

Identified high-engagement app categories through grouped statistics.

💰 Free vs Paid App Analysis

Compared pricing trends and installs across Free and Paid apps.

Investigated if higher price correlates with better ratings or fewer downloads.

📈 Feature Engineering & Correlation Insights

Created new features such as app size in MB and categorized installs.

Performed correlation analysis between size, ratings, reviews, and install ranges using heatmaps and pairplots.

🧠 Aggregation with GroupBy

Aggregated key metrics by app category and content rating to extract structured insights from noisy data.

📌 Full Code Available on GitHub: https://github.com/ifahadhashmi/Pandas-EDA-Mastery

🔍 Mastering EDA with YData Profiling | Pakistan Population Dataset 🇵🇰📊
In this Part of my data exploration journey, I used YData Profiling (formerly pandas-profiling) to perform a comprehensive, automated EDA on the Pakistan Population Dataset. This approach provided a rich and visual overview of the dataset, helping identify key patterns, distributions, and potential data quality issues—within minutes.

Here’s what I covered:

📋 Loading the Dataset & Generating a Profile Report
Loaded the dataset using Pandas and created a profile report with just a few lines of code using ydata_profiling.ProfileReport(), streamlining the initial data audit process.

🔍 Inspecting Variable Types & Data Summary
Quickly identified numerical, categorical, boolean, and date variables, allowing a structured understanding of the dataset’s shape and schema.

⚠️ Detecting Missing Values & Duplicates
The profiling report highlighted columns with missing data and duplicate rows. This helped plan appropriate cleaning strategies, like imputation or removal.

📊 Exploring Distributions & Correlations
Used built-in histograms, correlation matrices, and scatterplots from the report to detect outliers, skewness, and potential relationships between features.

💡 Highlighting Low-Variance & Constant Columns
Flagged uninformative features automatically—columns with constant or near-constant values—so they can be dropped to simplify the dataset.

📈 Exporting Interactive HTML Report
Saved the full analysis as an interactive HTML file using .to_file(), making it easy to share or revisit without rerunning the code.

📌 GitHub Code & Report: https://github.com/ifahadhashmi/Pandas-EDA-Mastery/blob/main/ydata_profiling.ipynb

🔍 Mastering EDA with Pandas – Part 4 | Exploring the Pakistan Population Dataset 🇵🇰📊
In Part 4 of my EDA & Pandas practice series, I focused on cleaning and exploring the Pakistan Population Dataset to extract structured insights from raw census data. This phase emphasized a systematic approach to understanding and preparing data for analysis.

Here’s what I covered:

🔢 Checking Data Types & Dataset Summary
Began by inspecting the structure of the dataset using .dtypes, .info(), and .describe(include='all').T to understand the format and spot any type mismatches.

🔍 Identifying Missing & Unique Values
Used .isnull().sum() to detect missing entries and .nunique() to explore the uniqueness of each column, helping ensure consistency across categories like districts and provinces.

📊 Visualizing Data Quality & Distributions
Plotted heatmaps to visualize missing values across the dataset. Utilized boxplots to examine the spread and identify anomalies, and histplots to understand the distribution of population data across different columns.

🧠 Combining Columns to Create a Total Population Feature
Created a new column AGE by summing six key fields (male, female, and transgender populations across rural and urban areas), providing a unified view of the total demographic per row.

🧮 Calculating Population Percentages
Calculated percentage increases by comparing current population columns against the 1998 baseline, offering insights into rural and urban growth trends.

📌 Aggregating Regional Insights with GroupBy
Used the groupby() function to summarize the dataset by DISTRICT and PROVINCE, making it easier to compare population metrics across different administrative regions.

⚙ Enhancing Readability with Display Settings
Used pd.set_option('display.max_columns', None) to view all columns clearly during the analysis, especially useful when working with wide census datasets.

📌 GitHub Code: https://github.com/ifahadhashmi/Pandas-EDA-Mastery/blob/main/Pandas_EDA_05.ipynb

🔬 Mastering Modern AI: 8 Specialized Model Architectures You Should Know 🧠🤖
As AI continues to evolve, we're seeing a shift from general-purpose models to task-optimized architectures that deliver better performance, speed, and scalability. Here's a quick breakdown of 8 specialized AI models and what makes each of them unique:

1. LLM – Large Language Model 💬
Purpose: Processes and generates human-like text.

How it works: Uses tokenization, embeddings, and transformers to understand and generate language.

Use cases: Chatbots, coding assistants, content generation (e.g., ChatGPT, Claude).

2. LCM – Latent Consistency Model 🎨
Purpose: Fast, high-quality image generation.

How it works: Uses latent diffusion with SONAR embeddings and quantization for efficient visual synthesis.

Use cases: Image generation (e.g., Stable Diffusion with LCM).

3. LAM – Language Action Model 🤖
Purpose: Enables perception-to-action pipelines in AI agents.

How it works: Integrates perception systems, intent recognition, and task planning.

Use cases: Robotics, AI agents that interact with their environments (e.g., Auto-GPT, Agent-1).

4. MoE – Mixture of Experts 🧑‍🏫🧑‍🔬
Purpose: Dynamically selects the best model "expert" for a task.

How it works: Uses a router mechanism to activate only relevant parts of the model for efficiency.

Use cases: Scalable AI systems with specialized sub-models (e.g., Google Switch Transformer).

5. VLM – Vision-Language Model 👁🗣
Purpose: Understands and reasons over both images and text.

How it works: Combines vision and text encoders through a multimodal processor.

Use cases: Image captioning, visual QA, multimodal search (e.g., CLIP, GPT-4V).

6. SLM – Small Language Model 📱
Purpose: Lightweight models for on-device and edge deployment.

How it works: Compact tokenization, efficient transformers, and memory optimization.

Use cases: AI on phones, wearables, or low-resource environments (e.g., Gemma, Phi-2).

7. MLM – Masked Language Model 🧩
Purpose: Understands context by predicting masked words.

How it works: Trains by masking words and predicting them based on bidirectional context.

Use cases: Pretraining for understanding tasks like classification, NER (e.g., BERT).

8. SAM – Segment Anything Model ✂📷
Purpose: Segment objects in images with high precision.

How it works: Uses a prompt/image encoder and mask decoder to isolate objects.

Use cases: Object detection, image editing, medical imaging (e.g., Meta's SAM).

🔍 Mastering EDA with Pandas – Part 2 | Deep Dive into Data Analysis 🧠📊

In this second part of my EDA & Pandas practice series, I explored essential data analysis techniques to better understand and visualize datasets.
Here’s what I covered:

📄 Descriptive Statistics & Data Summary
❓ Identifying Missing (Null) Values
🔥 Visualizing Data with Heatmaps
➕ Calculating Mean, 📈 Median, 🔁 Mode
🧬 Exploring Unique Values
📂 Selecting Specific Columns
👥 Grouping Data using groupby()
🔗 Revealing Patterns with Pairplots

📌 GitHub Code: https://github.com/ifahadhashmi/Pandas-EDA-Mastery/blob/main/Pandas_EDA_02.ipynb