Manav Thakur

Machine Learning Intern

New Delhi, IndiaData Science and Artificial Intelligence
1roles
25skills
1education
4credentials

About

Aspiring Data Analyst with hands-on experience in SQL, Python, and Power BI, delivering data-driven insights through exploratory data analysis, dashboarding, and predictive modeling on 10,000+ record datasets.

Experience

Machine Learning Intern

Unified Mentor Pvt. Ltd.

Apr 2026 – Jul 2026 · Gurgaon

Built a bank customer churn prediction and risk scoring system across 10,000+ client records, achieving 0.86 ROC-AUC by engineering six financial features and optimizing a Gradient Boosting model in Python and Scikit-learn. Identified behavioral engagement as the strongest churn predictor, finding a 3.8× lower churn rate among active versus inactive users. Built an interactive Streamlit dashboard for product line profitability analysis across 10,000+ sales transactions. Identified segments driving 80% of total profit using automated Pareto and Profit–Volume Quadrant Analysis algorithms.

Education

Guru Gobind Singh Indraprastha University (GGSIPU)

Bachelor of Technology, Artificial Intelligence & Machine Learning

Jul 2023 – Jul 2027

GPA: 9.1/10; Dean’s List; Top 10% of Cohort in all semesters; Relevant Coursework: Deep Learning, NLP, Machine Learning, DSA, SQL, Operating Systems

Skills

PythonSQLRegressionClassificationEnsemble MethodsFeature EngineeringModel EvaluationPandasNumPyExploratory Data AnalysisData CleaningData PipelinesLarge Language ModelsRetrieval-Augmented GenerationEmbeddingsChromaDBPrompt EngineeringOpenAI APIExcelScikit-learnStreamlitPower BIGitGitHub ActionsAWS

Projects

AI-Powered Resume Data Extractor

Engineered a low-latency GenAI extraction pipeline using the Google Gemini 2.5 Flash Lite API to parse unstructured resume text into ATS-ready JSON entities. Deployed an interactive Streamlit application for schema validation, parsed data outputs, and candidate summary metrics.

Real-Time Financial Fraud Detection System

Built an end-to-end machine learning pipeline processing five transaction types and engineering more than 10 custom features. Improved fraud scoring inference speed by 60% by optimizing pre-trained Joblib pipelines and deploying a Streamlit application for batch and real-time processing.