Ujjwal Singh Rao

Data Scientist & Machine Learning Engineer

GitHub LinkedIn Website

Passionate about AI, deep learning, and open source contributions. This portfolio showcases comprehensive Kaggle competition solutions demonstrating expertise across computer vision, NLP, and ensemble methods.

📁 Projects

22 comprehensive Kaggle competition solutions

Jigsaw Multilingual Toxic Comment Classification

🥇 Gold Medal

Ranked 3rd out of 1,621 participants. Multilingual solution using XLM-RoBERTa with adversarial training and two-stage model training pipeline.

NLP XLM-RoBERTa Adversarial Training Rank #3
Competition Documentation Code

SIIM-ISIC Melanoma Detection

🥇 Gold Medal

Ranked 6th out of 3,308 participants. Deep learning solution using EfficientNet with metadata fusion and advanced augmentation techniques.

Computer Vision EfficientNet Medical Imaging Rank #6
Competition Documentation Code

TalkingData AdTracking Fraud Detection

🥇 Gold Medal

Ranked 16th out of 3,943 participants. Large-scale fraud detection using LightGBM ensemble with sophisticated feature engineering.

Fraud Detection LightGBM Feature Engineering Rank #16
Competition Documentation Code

RSNA Intracranial Hemorrhage Detection

🥈 Silver Medal

Ranked 58th out of 1,345 teams. Multi-label classification of intracranial hemorrhages in CT scans using advanced CNN architectures.

Medical Imaging Computer Vision Medical AI Rank #58
Competition Documentation Code

Cassava Leaf Disease Classification

📊 Project

Ranked 589th out of 3,900 teams. Agricultural AI solution for identifying disease types on Cassava Leaf images.

Computer Vision Agriculture Deep Learning Rank #589
Competition Documentation Code

Humpback Whale Identification

📊 Project

Ranked 284th out of 2,120 teams. Fine-grained classification for identifying whales by their tail patterns.

Computer Vision Fine-grained Classification Deep Learning Rank #284
Competition Documentation Code

Avito Demand Prediction Challenge

🥈 Silver Medal

Ranked 21st out of 1,868 teams. Predict demand for online classified ads using ensemble methods and feature engineering.

Business ML Ensemble Feature Engineering Rank #21
Competition Documentation Code

Google QUEST Q&A Labeling

🥈 Silver Medal

Ranked 21st out of 1,571 teams. Improve automated understanding of complex question answer content using NLP techniques.

NLP Question Answering Text Analysis Rank #21
Competition Documentation Code

APTOS 2019 Blindness Detection

🥈 Silver Medal

Ranked 75th out of 2,928 teams. Detect diabetic retinopathy to prevent blindness using computer vision and medical AI.

Medical Imaging Computer Vision Medical AI Rank #75
Competition Documentation Code

Toxic Comment Classification Challenge

🥈 Silver Medal

Ranked 29th out of 4,539 teams. Identify and classify toxic online comments using NLP and text classification techniques.

NLP Text Classification Deep Learning Rank #29
Competition Documentation Code

Quick, Draw! Doodle Recognition Challenge

📊 Project

Ranked 138th out of 1,309 teams. Identify doodles accurately using computer vision and sketch recognition techniques.

Computer Vision Sketch Recognition Deep Learning Rank #138
Competition Documentation Code

iMet Collection 2019 - FGVC6

🥉 Bronze Medal

Ranked 61st out of 521 teams. Recognize artwork attributes from The Metropolitan Museum of Art using computer vision.

Computer Vision Cultural Heritage Deep Learning Rank #61
Competition Documentation Code

Tweet Sentiment Extraction

🥉 Bronze Medal

Ranked 208th out of 2,225 teams. Extract support phrases for sentiment labels using NLP and sentiment analysis.

NLP Sentiment Analysis Text Extraction Rank #208
Competition Documentation Code

Rainforest Connection Species Audio Detection

🥉 Bronze Medal

Ranked 78th out of 1,143 teams. Automate detection of bird and frog species using audio processing and CNN.

Audio Processing CNN Species Detection Rank #78
Competition Documentation Code

TGS Salt Identification Challenge

🥉 Bronze Medal

Ranked 292nd out of 3,219 teams. Segment salt deposits beneath Earth's surface using computer vision and segmentation.

Computer Vision Segmentation Geological Rank #292
Competition Documentation Code

Spooky Author Identification

📊 Project

Ranked 61st out of 1,241 teams. Identify horror authors from their writings using text classification and NLP.

NLP Author Identification Text Classification Rank #61
Competition Documentation Code

Statoil Iceberg Classifier

📊 Project

Ranked 336th out of 3,330 teams. Ship or iceberg detection from satellite images using CNN architectures.

Computer Vision CNN Satellite Imagery Rank #336
Competition Documentation Code

Toxic Comment Classification

📊 Project

Multi-label toxic comment classification using LSTM and CNN architectures with word embeddings.

NLP LSTM CNN Word Embeddings
Documentation Code

Gendered Pronoun Resolution

🥈 Silver Medal

Ranked 41st out of 838 teams. Pair pronouns to their correct entities using NLP and coreference resolution.

NLP Coreference Resolution Pronoun Resolution Rank #41
Competition Documentation Code