AI/ML Developer & Data Scientist
Nikhil Kumar
AI/ML Developer — Final-year BCA student specializing in Machine Learning and Generative AI, with hands-on experience in Python, data analysis, and developing practical AI solutions for real-world problems.
Scroll to explore
Core Capabilities
Programming Languages
HTML
CSS
Python
SQLAI/ML Frameworks
Scikit-learn
TensorFlow
Keras
Hugging FaceNLTKData & Visualization
NumPy
Pandas
Matplotlib
Seaborn
Power BI
TableauTools & Platforms
Jupyter Notebook
Google Colab
VS Code
Git
GitHub
StreamlitFeatured Projects
01
Credit Card Fraud Detection
- Trained & compared 4 models — Logistic Regression, Decision Tree, Random Forest & KNN — on 284,807 real transactions
- Best model reached 93.4% accuracy, 97.8% ROC-AUC with the lowest overfitting
- Handled extreme class imbalance (~0.17% fraud) via undersampling, validated on the full imbalanced holdout set (89.8% recall)
- "Fraud Shield" Streamlit dashboard — fraud-probability gauge, risk-factor breakdown & real transaction sampling
02
Employee Attrition Predictor
- Trained & compared 4 models — Logistic Regression, Decision Tree, Random Forest & Gradient Boosting — on IBM's 1,470-employee HR dataset
- Engineered custom features — TenureRatio, PromotionGap, SatisfactionScore, IncomePerYear — and balanced classes with SMOTE
- Best model (Logistic Regression) reached 86.4% accuracy, 82.9% ROC-AUC
- Deployed a live "AttritionSense" Streamlit dashboard with a risk gauge & automated HR action-plan recommendations
03
Resume IQ — Category Predictor
- NLP resume classifier trained on 2,484 resumes across 24 job categories using TF-IDF (15K features, bigrams)
- Compared 3 models — Logistic Regression, Linear SVM & Naive Bayes — best auto-selected via 5-fold CV, ~99% F1-score
- Text pipeline: cleaning, stopword removal & lemmatization (NLTK)
- Offline "ResumeIQ" Streamlit dashboard — PDF upload, top-3 predictions with confidence scores
04
Bank Customer Churn Prediction
- Trained & compared 3 models — Logistic Regression, Decision Tree & Random Forest — on 10,000 bank customers (France, Germany, Spain)
- Best model (Random Forest) reached 86.6% accuracy, 84.7% ROC-AUC
- EDA uncovered Germany's higher churn rate & inactive members as key risk factors
- Live "ChurnSense" Streamlit dashboard with a risk meter & banker action plan
Certifications
Machine Learning with Python
IBM
Data Analysis with Python
IBM
SQL and Relational Databases
IBM
Introduction to Generative AI
Google Cloud
Generative AI Foundations
upGrad with Microsoft
Finalist – FinCortex: AI CFO OS
Hackathon - IIT Roorkee · COMET'26
Generative AI Essentials
TCS iON Career Edge
IBM Dev Day: Bob Edition
IBM BOB Hackathon
Generative AI Mastermind
Outskill
Fundamental Course on Agentic AI
Microsoft
AI Engineer
OneRoadmap
Work History.
Building, breaking, and learning with passion.
Dec 2025 — Mar 2026
Data Science & Analytics
— Graphura India Pvt. Ltd. Internship
Built ML models and interactive dashboards on real-world datasets, focusing on performance optimization and deployment.
Jul 2025 — Oct 2025
Junior Data Scientist
— Eagletfly Solutions Internship
Designed and trained ML & deep learning models, working on data pipelines, model optimization, and real-world applications.
Let's Build
Something Great.
Have a project in mind or just want to connect? Drop me a message.