Hi 👋, I'm Mrityunjay Pathak

I'm a Data Scientist with a knack for uncovering patterns and trends that drive smarter decisions.

🎯 Tools and Technologies

• Programming Language : I'm familiar with Python, a powerful language for data science and machine learning.

• Libraries : I'm also familiar with essential data science libraries like NumPy, Pandas, Matplotlib, Seaborn and Plotly.

• Machine Learning : I have experience with Scikit-learn, a famous machine learning library used widely across industries.

• Database : I can work with MySQL, a popular database management system to handle and retrieve data effectively.

• BI Tool : I'm familiar with Power BI to perform data analysis, create dynamic dashboards and extract meaningful insights.

• Web Framework : I have experience with FastAPI, a high-performance web framework for building APIs with Python.

• Containerization : I can work with Docker for packaging application and their dependencies into containers.

• Version Control : I'm familiar with Git, which helps in keeping track of changes in code and collaborating effectively with a team.

📫 Connect with Me

Kaggle  |  LinkedIn  |  GitHub  |  Medium  |  Portfolio

➔ Problem

• In the used car market, buyers and sellers often struggle to determine a fair price for their vehicle.
• This project aims to provide accurate and transparent pricing for used cars by analyzing real-world data.
• It will assist both buyers and sellers make data-driven decisions and ensure fair transactions.

➔ Solution

To address this problem, I built and deployed a complete end-to-end machine learning pipeline :

• Data Collection
  • Scraped a dataset of 2,800+ used cars from Cars24 using Selenium and BeautifulSoup.
• Data Optimization
  • Optimized memory consumption of dataset by downcasting data types.
  • Stored the dataset in Parquet format, which compresses data without losing information.
  • It also provides much faster read/write speeds compared to CSV.
• Preprocessing & Modeling
  • Implemented Scikit-learn Pipelines & ColumnTransformer to prevent data leakage.
• API Deployment
  • Deployed the machine learning model as an API using FastAPI, with :
    • /predict endpoint for real-time predictions.
    • /health endpoint for monitoring API status.
    • Input validation & rate limiting for reliability.
• Frontend Integration
  • Designed a HTML/CSS/JS website to send API calls and display predictions in a user-friendly way.
• Containerization
  • Created a multi-stage Dockerfile with .dockerignore for building an optimized and lightweight Docker image.

➔ 𝗜𝗺𝗽𝗮𝗰𝘁

• Built and deployed a complete machine learning pipeline as a FastAPI application.
• Reduced dataset memory usage by 90% through data type optimization and Parquet conversion.
• Delivered 30% lower MAE and 12% higher R2-Score compared to the baseline model.
• Improved model stability by 70%, ensuring more consistent and reliable predictions.

➔ Problem

• With the rise of streaming services, viewers now have access to thousands of movies across platforms.
• As a result, many viewers spend more time browsing than actually watching.
• This problem can lead to frustration, lower satisfaction and less time spent on the platform.
• Which can impact both the user experience and business performance.

➔ Solution

• A content-based movie recommender system built with clean and modular code with proper version control.
• It analyzes metadata of 5000+ movies to recommend top 5 similar titles based on a user selected input.
• The system uses techniques like CountVectorizer and CosineSimilarity to recommend similar movies.
• The project not only focuses on functionality but on building a clean and scalable solution.

➔ Impact

If this system gets scaled and integrated with a streaming service, this could :

• Reduce the time users spend choosing what to watch.
• Increase user engagement, watch time and customer satisfaction.
• Help streaming platforms retain users by offering better personalized content.

➔ Objective

• To analyze Netflix content data, uncovering valuable insights into how the platform evolves over time.

➔ 𝗦𝗼𝗺𝗲 𝗞𝗲𝘆 𝗙𝗶𝗻𝗱𝗶𝗻𝗴𝘀

Cleaned and analyzed dataset of 8000+ Netflix Movies and TV Shows.

• More than 60% of content on Netflix is rated for mature audiences.
  • Suggests that Netflix targets adult viewers to boost engagement and retention.
• More than 25% of Movies and TV Shows are released on 1st day of the month.
  • Shows a consistent release schedule, likely to align with subscription cycles.
• More than 40% of the content on Netflix is exclusive to United States.
  • Shows a strong focus on the U.S. market and content availability by location.
• More than 20% of the content on Netflix falls under the "Drama" genre.
  • Confirms that "Drama" is a key part of Netflix's content library.
• More than 23% of the content on Netflix was released in 2019 alone.
  • Indicates a major content push that year, possibly tied to growth or user acquisition goals.

➔ Objective

• To analyze Supermarket Sales data, identifying key factors for improving profitability and operational efficiency.

➔ 𝗦𝗼𝗺𝗲 𝗞𝗲𝘆 𝗙𝗶𝗻𝗱𝗶𝗻𝗴𝘀

Analyzed purchasing pattern of 9000+ customers of Supermarket.

• More than 15% of the products sold were Snacks.
  • Shows that Snacks are a convenient choice and a big source of revenue.
• More than 32% of the sales were occurred in West region of Supermarket.
  • Suggests that West region is a strong performing area as compared to others.
• Health and Soft drinks are the most profitable category in Beverages.
  • Shows that both type of drinks option sells well.
• November was the most profitable month contributing about 15% of the total annual profits.
  • Makes it an ideal time for running promotions and special offers.