Analytical problem solver passionate about implementing practical AI solutions. I combine technical knowledge with attention to detail,adaptability, and collaborative skills. Eager to apply my foundation in computer vision and predictive modeling while continuing to grow in an innovative team environment.
● Spearheaded development of bank marketing prediction system, engineered XGBoost classifier with comprehensive feature engineering, achieving 92.1% ROC-AUC and 94% accuracy across 40,000 customer records.
● Architected innovative HR talent ranking system, implementing multi- agent evaluation with TF-IDF vectorization and K-means clustering, achieving 47.1% identification rate for target profiles.
● Pioneered Computer Vision model for MonReader's digitization app, engineered custom CNN architecture with motion detection layers, achieving 96.4% precision and 89.3% F1 score while reducing training time by 45% through optimized preprocessing techniques.
● Devised genetic algorithm optimization system, implemented mutation rate tuning and cluster normalization, significantly improving ranking diversity and reducing bias in candidate evaluation.
● Engineered customer satisfaction prediction model, designed feature importance analysis and elimination techniques, attaining 65.4% accuracy with XGBoost while maintaining interpretability.
● Conceived automated supply chain risk analysis system, tasked withdeveloping proof-of-concept, engineered solution implementing ReActpattern with LangChain and Groq LLM, successfully processing 1,700+news articles.
● Identified system optimization challenges, required to enhanceperformance, designed memory system with context management and ratelimiting, enabling efficient large-scale data processing within APIconstraints.
● Collaborated with CEO on system architecture, required structuredanalysis output, implemented chain of thought prompting for executivesummary generation, delivering actionable insights across risk categories.
Programming & Libraries: Python(Pytorch, NumPy, Pandas), SQL, C,C
Machine Learning: Scikit-learn(XGBoost, RandomForest, DecisionTree), K-means Clustering, GeneticAlgorithms
Statistical Analysis: Univariate/Bivariate Analysis,Confidence Intervals, CalibrationCurves, ROC-AUC Analysis
Feature Engineering: InteractionFeatures, Threshold Features,Aggregate Features, TF-IDFVectorization
Natural Language Processing: CosineSimilarity, FuzzyWuzzy StringMatching, LangChain, LLMIntegration, Chain-of-thoughtPrompting
Computer Vision: ThreadPoolExecutor ParallelProcessing, Data Augmentation(Rotation, Brightness, Shifts), RGBQuality Validation
System Design: Memory Management,Rate Limiting (RPM/RPD), LoggingSystems, Performance HistoryTracking
MonReader (current)
● Engineered a high performance Computer Vision application detecting page flips in real time with 96.4% precision, enabling seamless document digitization for visually impaired users.
●Designed an optimized image pre-processing pipeline with intelligent cropping and contrast enhancement, reducing training time by 45% while improving model accuracy.
●Developed a custom CNN architecture with specialized motion detection layers that achieved 89.3% F1 score on 2,989 images from 130 unique videos.
●Implemented CPU optimized data processing techniques reducing inference time to 12.5 minutes for the entire dataset, enabling responsive real-time detection.
●Analyzed model performance through comprehensive metrics tracking and confusion matrix visualization, achieving 97.5% confidence on positive detections.
HR Talent Ranking System
● Architected ML system with multi-agent evaluation, integrating TF-IDFvectorization, K-means clustering(n=5), and genetic algorithms (population=50, generations=10,mutation=0.1).
● Engineered scoring system with weighted agents (Title 90%, Location 5%, Network 5%), implementing FuzzyWuzzy for string matching and cosine similarity for cluster analysis.
● Developed sophisticated bias prevention through cluster normalization and genetic optimization, with comprehensive fitness scoring and crossover implementations.
● Constructed logging system with performance history tracking, enabling detailed model diagnostics and iterative improvements.
Bank Marketing Campaign Predictor
● Engineered ML solution for term deposit prediction, processing imbalanced dataset (92.8% no, 7.2% yes) with univariate/bivariate analysis.
● Implemented XGBoost with probability calibration and stratified 5-fold cross-validation, achieving 92.1% ROC-AUC and 94% accuracy.
● Developed comprehensive feature engineering incorporating interaction features, threshold features, and aggregate features.
● Generated actionable insights through calibration curves and duration pattern analysis (15-30 mins: 63% success).
Customer Happiness Predictor
● Developed end-to-end ML solution for customer satisfaction prediction, engineered comprehensive feature set including interaction scores and threshold features, achieving 65.4%accuracy using XGBoost.
● Conducted extensive exploratory analysis, implementing correlation studies, univariate distributions, and bivariate feature analysis, revealing key satisfaction drivers across six service dimensions.
● Created advanced feature engineering pipeline, developing interaction metrics (service, value, experience scores)
and threshold features, significantly improving model interpretability.
● Implemented model comparison framework, analyzing predictions vs observed patterns and feature importance
across Decision Tree, Random Forest, and XGBoost models, enabling data driven selection of the optimal model.
● Data Vidhya Python for DataEngineering Certificate
● Data Vidhya SQL for DataEngineering Certificate
● ML Scientist Career Track(DataCamp) - In Progress
● AI Agents Course By Hugging Face -In Progress
English, Malayalam, Kannada and Hindi.
Playing football, hiking and listening to podcasts