Saikat Mandal
Data science / ML
Software dev projects
I'm a Software Engineer crafting powerful web & mobile apps with code, creativity, and coffee.
Find me on Twitter or on LinkedIn, or just send me an email saying hi
Professional Work
TCS (Postnord)
On site, Pune India · Jul 2024 - Present
Software Engineer
• Built RESTful APIs in Java using Spring Boot, with Hibernate for PostgreSQL database interactions
• Designed and maintained automated end-to-end (E2E) tests with Postman, alongside integration and unit tests using JUnit
• Wrote unit and integration test cases for React applications using React Testing Library and Vitest, ensuring frontend reliability.
• Developed reusable React components to enhance UI consistency and maintainability.
•Dockerized and deployed applications, streamlining development workflows and improving deployment efficiency.
Qvik.io
Remote, Noida India · Feb 2023 – May 2023
Software developer intern
• Built RESTful APIs in Java using Spring Boot, with Hibernate for PostgreSQL databaseDeveloped an intuitive React.js platform with responsive UI using Tailwind CSS, aligned with Figma designs.
• Implemented JWT-based authentication for secure user authorization.
• Collaborated within an agile team environment, mentoring new interns to enhance team productivity and knowledge sharing.
Data science / Machine learning projects
Classification problems
Heart Disease Prediction is a classic binary classification problem in machine learning. Using datasets like the UCI Heart Disease dataset, I built a model to predict the likelihood of heart disease based on medical attributes such as age, cholesterol, and blood pressure. This project improved my skills in feature engineering and highlighted the real-world impact of ML in healthcare.
Iris Flower Classification is a well-known multiclass classification problem in machine learning. Using the classic Iris dataset, I built a Logistic Regression model from scratch with NumPy and math to classify iris species based on features like petal length, petal width, sepal length, and sepal width. This project strengthened my understanding of core ML algorithms, especially how multiclass classification can be handled using a One-vs-Rest approach, and gave me hands-on experience in implementing machine learning logic without any external libraries.
Titanic: Machine Learning from Disaster is a knowledge competition on Kaggle. Many people started practicing in machine learning with this competition, so did I. This is a binary classification problem: based on information about Titanic passengers we predict whether they survived or not. General description and data are available on Kaggle. Titanic dataset provides interesting opportunities for feature engineering.
Custom models
I built a Linear Regression model from scratch using only NumPy and Python's built-in math module, without relying on machine learning libraries like scikit-learn. The model was implemented using the Gradient Descent algorithm to optimize the weights by minimizing the mean squared error. I manually handled all the core components, including predictions, loss calculation, gradient computation, and model training. This project helped me deeply understand how linear regression works under the hood, including the role of learning rate, the effect of iterations, and how convergence happens over time.
I implemented a Logistic Regression model from scratch using NumPy and the math module in Python, without using any machine learning libraries. I manually coded the sigmoid activation function, binary cross-entropy loss, and the Gradient Descent algorithm to optimize the model's parameters. The model was trained to classify binary outcomes by learning from labeled data, and I added functions to predict probabilities and final class labels. This project gave me a clear understanding of the mathematical foundation behind logistic regression, including how it models probabilities and updates weights through backpropagation.