I am a final-year B.E. Artificial Intelligence & Machine Learning student at Poojya Doddappa Appa College of Engineering, Kalaburagi, Karnataka, graduating in 2027.
I design and ship end-to-end machine learning systems — from raw data ingestion and feature engineering through model training, evaluation, and Streamlit-based production deployment. My work spans customer analytics, dynamic pricing, NLP pipelines, and supply chain risk modelling.
Alongside ML, I am building full-stack web development competencies with HTML, CSS, JavaScript, Django, and SQL/MongoDB, combining backend engineering discipline with data-driven product thinking.
Open To:
- Machine Learning Internships
- Data Science Internships
- Full Stack Development Internships
- Research Collaborations in AI / NLP
Languages
Frontend & Web
Backend, Databases & ML
ML / Data Science Ecosystem
Dev Tools & Cloud
| Domain | Proficiency | Details |
|---|---|---|
| Supervised Learning | Advanced | Linear / Logistic Regression, Random Forest, XGBoost, CatBoost, SVM |
| Feature Engineering | Advanced | Pipeline design, encoding, scaling, temporal & behavioral features |
| NLP / Text Analytics | Intermediate | TF-IDF, tokenization, stemming, lemmatization, semantic similarity |
| Model Evaluation | Advanced | Accuracy, Precision, Recall, F1, ROC-AUC, RMSE, R² scoring |
| Hyperparameter Tuning | Intermediate | GridSearchCV, cross-validation, performance benchmarking |
| Data Preprocessing | Advanced | Missing value handling, outlier detection, normalization, encoding |
| ML Deployment | Intermediate | Streamlit dashboards, modular src/ pipeline architecture |
| Explainable AI | Intermediate | Bias detection, transparent scoring, SHAP awareness |
| Data Visualization | Intermediate | Plotly interactive charts, Matplotlib, Seaborn statistical plots |
| Cloud / Azure NLP | Beginner–Intermediate | Natural Language Solutions in Azure (Microsoft certified) |
RoleFit AI — NLP-Powered Resume Screening System
An intelligent hiring assistant that evaluates candidate resumes against job descriptions using Natural Language Processing and Explainable AI techniques. Ranks candidates by composite fit score, detects skill gaps, identifies biased language in job descriptions, and auto-generates structured interview questions.
| Attribute | Details |
|---|---|
| Stack | Python, spaCy, NLTK, scikit-learn, TF-IDF, Plotly, Streamlit, PyPDF2 |
| Scale | Multi-resume batch processing with leaderboard ranking |
| Performance | Hybrid TF-IDF + cosine similarity scoring with skill-coverage weighting |
| Security | Responsible AI — bias detection in job descriptions, demographic-neutral scoring |
| Impact | Decision-support tool for HR automation; reduces manual screening time significantly |
| Repository | github.com/UMESH-KALE0777/rolefit-ai |
The system implements a modular five-stage pipeline: PDF text extraction → NLP preprocessing → skill extraction → hybrid candidate scoring → Explainable AI dashboard. Applied text preprocessing techniques including tokenization, stop-word removal, and stemming/lemmatization, then trained and benchmarked Naive Bayes, Logistic Regression, and SVM classifiers for optimal classification performance. The radar chart analytics and multi-candidate ranking leaderboard make shortlisting decisions transparent and auditable for recruiters.
Fashion Price Intelligence — AI-Powered Dynamic Pricing Dashboard
A machine learning-powered pricing and business intelligence platform that predicts optimal selling prices for fashion products and generates actionable retail analytics. Bridges predictive modelling and real-world retail strategy for fashion startups and pricing analysts.
| Attribute | Details |
|---|---|
| Stack | Python, CatBoostRegressor, Pandas, NumPy, scikit-learn, Plotly, Streamlit |
| Scale | 25,000 rows (20k train / 5k test), modular src/ production architecture |
| Performance | R² Score: 0.97 — highly accurate pricing predictions with stable variance |
| Security | Input validation, clean separation of training artefacts from live inference |
| Impact | Enables data-driven pricing decisions; identifies bestsellers and brand strength indices |
| Repository | github.com/UMESH-KALE0777/fashion-price-intelligence |
Implemented systematic hyperparameter tuning using GridSearchCV to significantly improve baseline model performance. The automated feature engineering pipeline processes product attributes, seasonal trends, and market signals. An interactive Streamlit dashboard delivers real-time price predictions alongside category performance breakdowns, brand strength indices, and strategic pricing recommendations — making complex ML outputs immediately actionable for non-technical stakeholders.
Customer Churn Prediction — Binary Classification System
An end-to-end binary classification system predicting customer churn for business retention strategy. Built a production-grade preprocessing pipeline from raw customer behavioral data through to real-time Streamlit deployment.
| Attribute | Details |
|---|---|
| Stack | Python, XGBoost, Random Forest, scikit-learn, Streamlit, Pandas |
| Scale | Real-world customer behavioral dataset with full preprocessing pipeline |
| Performance | 85%+ accuracy; evaluated on Accuracy, Precision, Recall, F1, ROC-AUC |
| Security | Input sanitization; model artefacts versioned and separated from application layer |
| Impact | Real-time churn risk assessment dashboard for business stakeholders |
| Repository | github.com/UMESH-KALE0777 |
Engineered a complete feature engineering pipeline extracting meaningful signals from raw behavioral records. Benchmarked XGBoost against Random Forest, selecting the optimal model via cross-validated ROC-AUC scoring. The deployed Streamlit application enables business users to input customer attributes and receive an instant churn probability score — translating ML predictions into directly actionable retention decisions.
Student Marks Prediction — ML Regression Pipeline
A clean, beginner-to-intermediate machine learning regression project predicting student academic performance from study habits and engagement metrics. Structured using industry-grade modular Python architecture.
| Attribute | Details |
|---|---|
| Stack | Python, scikit-learn, Linear Regression, Streamlit, Pandas |
| Scale | Structured dataset covering study hours, attendance, and assignment completion |
| Performance | Reproducible model training with persisted .pkl artefacts |
| Security | Modular src/ separation; no hardcoded paths or data leakage |
| Impact | Interactive Streamlit web app for real-time mark prediction |
| Repository | github.com/UMESH-KALE0777/student-marks-prediction |
Demonstrates a complete, reproducible ML workflow: data ingestion → preprocessing → model training → artefact serialization → Streamlit UI deployment. Built as a reference-grade project showcasing clean pipeline architecture and deployment-ready code organisation.
Supply Chain Risk Analysis — Predictive Risk Intelligence
A data-driven supply chain risk assessment system applying ensemble classification to predict disruption probability from operational datasets.
| Attribute | Details |
|---|---|
| Stack | Python, scikit-learn, Random Forest, Logistic Regression, Pandas, Seaborn |
| Scale | Multi-feature supply chain operational dataset |
| Performance | Actionable confidence scores per risk prediction |
| Security | Clean EDA-to-modelling pipeline with no data contamination |
| Impact | Strategic insights and visualizations supporting supply chain decision-making |
| Repository | github.com/UMESH-KALE0777 |
Conducted thorough EDA to identify key disruption factors and risk patterns. Applied Random Forest and Logistic Regression classifiers to deliver per-record risk probability scores. Generated executive-level visualizations enabling supply chain teams to prioritize mitigation efforts on highest-probability disruption events.
Machine Learning with Python Intern — EDXcellence LLP
Jan 2026 – Mar 2026 · 3 months · Remote, Bangalore, Karnataka
Completed a structured 2-month online internship in Machine Learning with Python, delivering production-ready ML project work under industry mentorship.
- Designed, trained, and evaluated supervised ML models using Python and scikit-learn
- Applied XGBoost gradient boosting for classification and regression tasks with real-world datasets
- Implemented end-to-end data preprocessing, feature engineering, and model evaluation pipelines
- Received Outstanding Performance Award for dedication, problem-solving, and exceptional technical delivery
| Recognition | Details |
|---|---|
| Outstanding Performance Award | Awarded by EDXcellence LLP for exceptional achievement during the Machine Learning with Python internship (Apr 2026) |
| NextGen AI Buildathon | Participated in a competitive hackathon; developed and presented a time-constrained AI-based solution |
| LeetCode Problem Solver | 39+ problems solved — Arrays, Strings, Hash Tables, Dynamic Programming, Sliding Window |
| 103+ Learning Hours | Completed 103+ hours of structured certification coursework across Python and Data Science |
EDXcellence LLP
Saylor University
Microsoft
Cisco
Forage
Edunet Foundation
National Skill Development Corporation
current_focus:
learning:
- Deep Learning fundamentals (Neural Networks, CNNs, RNNs)
- Django REST Framework for production API development
- Azure Machine Learning Services
- Advanced NLP with transformer-based models
building:
- NLP-powered HR automation tools
- Full-stack Django + ML integrated web applications
- End-to-end ML pipelines with modular production architecture
exploring:
- LLM integration into real-world product workflows
- BERT and sentence transformers for semantic search
- Cloud deployment of ML models (Azure, Streamlit Cloud)
open_to:
- Machine Learning Internships (Remote / Hybrid)
- Data Science Internships
- Full Stack Development Internships
- AI Research Collaborations
- Open Source ML Projects
"Build systems that learn. Engineer solutions that scale. Ship work that matters."