Seetharaman Radhakrishnan
MSc Artificial Intelligence| 2x AWS Certified | Data Scientist | Python Developer | AI/ML Engineer | Open to Relocation

Meal Recommender

Prototyped an AI-powered meal planning assistant that scrapes real recipes via web scraping nd parses cook times for user-defined time filtering. Implemented an interactive UI using Streamlit and integrated Ollama’s local LLM (Gemma 3B) via LangChain for intelligent and personalized meal selection.

End to End complete data pipeline including scraping, time parsing, and dynamic filtering. Utilized a LangChain Pandas agent to generate daily meal plans (breakfast, lunch, and dinner) by randomly selecting from filtered results. The system reduced user decision fatigue by over 60%, streamlining the meal selection process for time-constrained individuals.

Superstore Sales Analysis

Architected an interactive analytics dashboard and data pipeline leveraging Streamlit, Pandas, Seaborn, and scikit-learn to analyze historical retail sales data. Implemented KMeans clustering for customer segmentation, evaluated clusters using F1-score and precision, and visualized them using PCA. Enabled dynamic filtering by year and region through real-time KPIs and charts.

Student Performance Analysis

Analyzed academic performance using student data from the UCI Machine Learning Repository. Implemented a structured data analysis pipeline involving data cleaning, exploratory data analysis (EDA), and visualization to identify key factors influencing final grades (G3).

Applied Random Forest Regression and Logistic Regression models to assess predictors, highlighting G2 (second period grade) as the most significant. Improved model accuracy from 91% with Logistic Regression to 92% with Random Forest. Additional factors such as study time, parental education, and alcohol consumption were also examined.

Ipl 2025 Analysis

Analyzed ball-by-ball data from the IPL 2025 season to extract insights on player performances, team statistics, and overall tournament trends. Performed data cleaning and exploratory data analysis on over 17,000 deliveries, handling missing values related to wickets and fielding.

Explored dismissal types, top batsmen and bowlers, team-wise runs, and boundary hitting patterns, with a special focus on Chennai Super Kings. Visualized key findings using Seaborn and Matplotlib in a structured Jupyter Notebook workflow.

Math Utility Toolkit

The Math Utility Toolkit is a modular Python application that integrates essential mathematical tools, including a basic calculator, geometry helper, statistical functions, quadratic equation solver, and sequence generators (arithmetic, geometric, Fibonacci). It offers both a command-line interface and an interactive Streamlit GUI.

Designed using Object-Oriented Programming principles, the project emphasizes clean, maintainable, and scalable code with approximately 1000 lines. Functionality was rigorously validated through over 25 test cases, ensuring reliability and enhanced user accessibility.

Basic math utility Toolkit

Designed and implemented a modular, command-line Python application offering a suite of everyday tools using object-oriented programming. The toolkit includes a unit converter (temperature, weight, distance), a Caesar cipher encryption/decryption tool, and various mini-games like dice roller, number guessing, lottery simulator, and coin toss.

It also features a scientific calculator with trigonometric, logarithmic, and exponential functions, a matrix calculator (with NumPy integration) for addition, subtraction, and transposition, and a date utility to calculate age, birthday countdowns, and time differences. The project emphasizes modular class design, code reusability, and interactive user experience.

Schwa Identification

Collaborated with a client to implement an AI-based speech analysis pipeline focused on identifying schwa-related pronunciation challenges in ESOL learners.

Processed recorded speech data to extract granular phonetic features such as frequency, duration, intensity, and spectral characteristics specific to the schwa sound. Shifted from word-level to phoneme-level feedback, allowing for more precise and personalised pronunciation insights. Used signal processing techniques to detect subtle deviations from standard British English pronunciation.

Designed the pipeline to be modular and scalable, enabling continuous algorithm refinement and easy extension to other phonetic targets.

Heart disease prediction

Implemented a machine learning prototype to predict individual heart disease risk using Python libraries such as scikit-learn, NumPy, Pandas, and Seaborn. Applied Random Forest and K-Nearest Neighbors algorithms, gaining hands-on experience in model tuning and performance optimization.

The project strengthened skills in data preprocessing, exploratory analysis, and predictive modeling, with a focus on health-related datasets and real-world applicability.