Vinay
Chanamallu
Data Scientist & ML Engineer
I transform data into intelligent solutions through end-to-end ML pipelines, advanced feature engineering, and production-ready models that deliver measurable business impact.
Turning Data Into
Intelligent Solutions
Transforming complex data challenges into scalable, production-ready machine learning solutions.
I'm Venkata Chandrasekhar Vinay Chanamallu, a data scientist with 2.5+ years of experience turning messy business problems into measurable outcomes. I've worked across fraud, churn, and customer-support automation, building models, validating them carefully, and partnering with operations and engineering to ship changes that actually get used.
At EPAM Systems, I tuned a LightGBM classifier with Optuna and isotonic calibration on 2.3M transactions, reducing weekly false declines from 520 to 455 at recall 0.80. I built an XGBoost fraud model that lifted legit approvals from 89% to 94% in a 6-week A/B test on 120k transactions. I also built a churn-risk model achieving 7% relative reduction in churn and auto-routed 120k Zendesk tickets using TF-IDF classification, cutting first-response time by 8%.
My expertise spans LightGBM, XGBoost, CatBoost, logistic regression, isotonic calibration, and threshold/cost analysis with Optuna optimization. I'm experienced in A/B testing, time-based splits, ROC/PR evaluation, SHAP explainability, and production monitoring with MLflow drift detection (PSI, score distributions).
Recently at HealID, I owned the mobile ingestion layer for 40+ HealthKit/Fitbit metrics, co-designed a GCP pipeline (Cloud Run, Pub/Sub, GCS) handling 10,000 events/day with under 1% ingest errors, and partnered to serve a LightGBM readiness model with p50 inference of 250ms. Known for being practical: clear metrics, clean handoffs, and solutions that hold up in the real world.
Professional
Journey
My evolution from computer science student to ML engineer and business analyst.
Data Engineer Intern
HealID Inc.
MS Business Analytics
University of Maryland
Data Scientist
EPAM Systems
Junior Data Scientist
EPAM Systems
BTech Computer Science
JNTUK
My Tech Stack
Expertise across machine learning engineering and deployment technologies
Programming
3 skills
Python
SQL
Pandas
ML & Modeling
9 skills
XGBoost
LightGBM
CatBoost
scikit-learn
Gradient Boosting
Random Forest
Elastic Net
Optuna
SHAP
NLP & Time Series
4 skills
TF-IDF
ARIMA
Time Series
Prophet
Cloud & ETL
6 skills
AWS Lambda
AWS S3
AWS Redshift
AWS Glue
Athena
CloudWatch
Visualization
3 skills
Tableau
Plotly
Microsoft Excel
Projects
A few case studies that show how I design, build, and ship.
Medical Image Classification
Automated pneumonia detection from chest X-rays using deep learning and CNN transfer learning. Achieved 95.21% accuracy with EfficientNetB0 on 5,840 real chest X-ray images.
Airbnb High-Booking Prediction
Competition-winning model (3rd place) for identifying high-booking rate Airbnb listings. ROC AUC 0.91176 using stacked XGBoost, CatBoost ensemble with SHAP explainability.
Flight Delay Prediction Pipeline
End-to-end ML pipeline predicting flight delays from 3.6M records with weather enrichment. 70.44% accuracy with deployment-ready CLI tool.
CNN Image Classification
Custom CNN architecture for CIFAR-10 object classification with 10 classes. Achieved 82.45% test accuracy with data augmentation and comprehensive evaluation.
Customer Churn Prediction
Business-focused churn prediction model identifying $227K+ at-risk revenue. 0.705 ROC-AUC with Gradient Boosting and actionable business insights.
Breast Cancer ML Coach
Lightweight ML pipeline for breast cancer detection with 98% hold-out accuracy. Modular Python package with YAML configuration and automated cross-validation.
RAG Chatbot
Intelligent Retrieval-Augmented Generation chatbot leveraging advanced NLP techniques for contextually relevant responses.
Financial Market ML Analysis
Comprehensive stock prediction system with real-time Yahoo Finance data. Lasso Regression achieved 99.26% R² for price prediction across 10 major companies.
Time-Series Energy Forecasting
Advanced time series forecasting model for India's energy consumption using ARIMA, SARIMA, and LSTM. Predicts 7% demand increase over 5 years.
Movie Sentiment Analysis (LSTM)
Enhanced movie review sentiment analysis using LSTM neural networks. 90-95% training accuracy on IMDB 50k reviews with interactive demo and confidence scoring.
LLM 101: Bytes to GPT
Educational project building a tiny GPT model from scratch. Learn LLMs through tokenization, bigram models, and baby transformers with clear, commented code.
LSTM Action Recognition
Deep learning model using LSTM networks for human action recognition from video sequences. Bidirectional LSTM improved accuracy by 12%.
Hand Gesture Counter
Real-time hand gesture recognition system detecting and counting finger gestures (1-5) using MediaPipe with text-to-speech announcements.
Congressional Speech Analyzer
NLP analysis of U.S. Congressional speech data from 2020. Compares linguistic patterns between Democratic and Republican senators using spaCy and bigram analysis.
Doodle Digit App
Interactive handwritten digit recognition using k-NN classifier on 8x8 pixel images with Gradio web interface. Educational ML project with browser-based drawing.
E-Commerce Analytics Platform
Comprehensive SQL-based analytics system for multi-vendor e-commerce. Features RFM segmentation, cohort analysis, and churn prediction handling millions of transactions.
IAD Flight Departures EDA
Multi-dimensional exploratory analysis of flight departures at Washington Dulles Airport. Analyzes carrier performance, COVID impact, economic correlations, and weather delays.
Crash Duration Prediction
Predictive modeling system for estimating crash durations using machine learning techniques. Supports transportation and incident management applications.
Amtrak Performance Analysis
SQL-based analysis examining Amtrak's operational performance. Identified routes with highest delays and recommended buffer times reducing delays by 12%.
GET IN TOUCH
Ready to collaborate on your next data science project? Let's build something impactful together.
Send a Message
I typically respond within 24 hours