PRATHVIK G S | 20PH20028

PHYSICS (M.Sc. 5Y)

EDUCATION
Year Degree/Exam Institute CGPA/Marks
2025 M.SC(5YR) IIT Kharagpur 8.39 / 10
2020 CBSE-AISSCE Jawahar Navodaya Vidyalaya, Pondicherry 93.8%
2018 CBSE-AISSE Jawahar Navodaya Vidyalaya, Hassan 92.8%
INTERNSHIPS
Data Science Intern | TCG Digital, Kolkata, India | Semi-automated solution generation using GenAI May'24- Jul' 24
•Developed a semi-automated data preprocessing tool using Generative AI, LangChain’s Pandas agent, OpenAI API and prompt engineering for LLMs
•Implemented an intuitive, click-based interactive interface that enables efficient data cleaning and post-processing steps with minimal to no coding.
• Built the entire poc with an interactive user interface, integrating a chatbot powered by LangChain’s DataFrame agent and OpenAI API Utilizing GPT-4o.
•Optimized predictive Analytics workflows, transition time from raw data to predictive insights, decreasing time from 4-5 weeks to 1-2 days.
•Engineered a intuitive tool to generate dashboard UI codes and sample data quickly from images using Langchain with multimodal LLM (GPT-4o).
•Utilized LangChain to parse and structure the output in JSON Providing an integrated functionality for downloading the generated codes and data.
Research Intern | Max Planck Institute, Dresden, Germany |(Prof Ivo. F. Sbalzarini .Particle based PINNs) May'23 -Jul'23
•Used SPH Kernel to approximate functions and replaced automatic differentiation in the loss functions of Physics Informed Neural Networks(PINNs).
•Developed an algorithm by making efficient use of celllists for updating the position of the singularity during the training for non-smooth curves.
•This method reduced the training time of each epoch further. For 2500 data points, the computational time was reduced by more than 10 folds.

PROJECTS
Stock Market Analysis and Price Prediction Using LSTM | Self project Jun'24-Jun'24
•Conducted analysis of stock market trends for major tech companies using time series, focusing on price changes, daily returns, and correlations.
•Implemented LSTM to predict future stock prices and visualized key metrics using seaborn, matplotlib etc. Extracted the data from yahoo finance
•Evaluated and quantified investment risks by calculating Value at Risk (VaR), a data driven approch to measure the risk of investing in a particular stock
Credit Card Fraud Analysis using Machine Learning | Classification problem Aug'22-Aug'22
•Analyzed over 6.5 Million rows of data using various ML models to detect fraudulent transactions. Obtained the best F1 score with Random Forest
•Handled the skewed data with majority undersampling,minority oversampling, synthetic data for minority data, etc. used appropriate metrics.
•Used 80:20 train test split. Obtained an F1 score of 87% and an accuracy of 99.97% on test data. Used various algorithms like bagging, boosting etc.
Product Recommendation engine for ecommerce| Self project Jul'24-Jul'24
•Implemented a system recommending popular products to new users, addressing the cold start problem and providing immediate value.
•Developed a collaborative filtering model that user to offer personalized product recommendations based on users history and rating by similar users.
•Developed a system for products without user-item purchase history, utilizing textual clustering of products descriptions for suggestions based on search.
Bachelors Thesis Project | Deep Operator Networks for PDEs Jan'24-Apr'24
•Implemented Deep Operator Network Architecture capable of learning Operators using Tensorflows’ functional API for reaction diffusion equation.
•Implemented custom Physics informed loss functions using equations for training the Deeponet network. Fine tuned hyperparameters like learning rate
•Trained the network for various Equations, validated accuracy with the numerical solutions , obtained avg R2 score of 0.95+ on new test functions

COMPETITION/CONFERENCE
Loan Eligibility Prediction | Binary Classification problem | Analytics Vidhya Hackathon Mar'23-Mar'23
•Ranked 2089 of 79,000 participants. Predicted loan eligibility based on features like income, loan amount, credit history, education and other features.
•Used the dataset to identify the customer segments to be targetd. Used logistic Regression, KNN , Decision Trees , random forest etc for classification
•Performed data preprocessing steps like standardizing the data, data-splitting for training and testing, missing value handling and feature selection.
•Used metrices like accuracy, precision, recall, F1 score and confusion matrix on test data to evaluate different algorithms for the best model selection.
Prediction of number of upvotes | Regression problem | Analytics Vidhya Hackathon Jul'22-Jul '22
•Identifying the best authors on an online platform by predicting the number of upvotes based on features like views, reputation etc.
•Used Linear regression, Deep Neural Networks and other ML models in the model to predict the number of upvotes. This prediction and identification
brings insight into increasing the user engagement of the platform. Ranked 577 out of 6,000 registered participants.

COURSEWORK INFORMATION
•Academic courses: Advanced Calculus | Programming and Data Structures | Probability and Statistics ,Linear Algebra and Complex Analysis | High
Performance computing and its application in complex systems |Essentials of scientific computing | Computational physics Lab
•Online Courses : Machine learning specialization Coursera, LangChain for LLM Application Development , ChatGPT Prompt Engineering for Developers-
coursera, LangChain Chat with Your Data -deeplearning.AI| Neural Networks and Deep learning - Coursera| Machine Learning A-z-Udemy

SKILLS AND EXPERTISE

Technical Skills: Programming Languages : C, C++, Python, SQL| Libraries: Tensorflow, Langchain, scikit-learn, Numpy, Pandas, Matplotlib, Standard
Template Library, Pycaret, Plotly, Streamlit, XGBoost, scipy, Openai Tools and softwares: Jupyter Notebook,Google Colab, VS Code,LaTex, Linux commands
Codeforces Rating - 1498 (Max), Specialist. Solved 500+ algorithmic problems applying knowledge of data structures and algorithms with critical thinking.

AWARDS AND ACHIEVEMENTS

• Selected for DAAD SECAI Scholarship for master thesis, TU Dresden, Germany
• INSPIRE Scholarship awardee by the Departmentof Science and Technology, Governmentof India.

EXTRA CURRICULAR ACTIVITIES

•Investment Enthusiast: Engage in fundamental analysis, stock market research and staying updated with financial news to make informed investment
decisions. Utilized stock screeners for effective stock screening based on various parameters. Achieved an XIRR of 50%+ on personal portfolio.
•Authored 22 blogs on recent advancements in AI and ML, keeping up withe latest technologies | Holder of NCC A Certificate and Judo Brown Belt.

!Self declared by the student, CDC could not verify the relevant documents