Krishna Balachandran Nair

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.

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