Data Scientist @ IndiGo | Agentic AI · LLMs · RAG · Generative AI · MLOps | Production ML Systems for Aviation | BITS Pi

Bhuman Pandita  • Peerlist

• Experimented with different CNN architectures, including ResNet, DenseNet, and MobileNetV2, and designed a new 6-layer model to classify 13 distinct sound classes. • Implemented preprocessing techniques using libraries such as librosa to generate Mel spectrograms, facilitating the CNN-based classification process

CNN-Based Sound Classification for Multi-Clas

• Developed a CycleGAN model for image dehazing, leveraging its ability to learn mappings between unpaired image domains while enforcing cycle consistency and adversarial training. • Explored the impact of data augmentation techniques on improving image quality and generalization. • Implemented the model from scratch without using any pre-trained weights, showcasing a deep understanding of the underlying principles.

Enhancing Image Clarity through CycleGAN-base

This project focuses on flood prediction in the state of Kerala. The objective is to develop machine learning models that can accurately predict the occurrence of floods based on various features, such as monthly and annual rainfall index. The project explores six algorithms: Logistic Regression K-Nearest Neighbors Decision Trees Random Forests Naive Bayes Support Vector Machines

Flood Warning System Using Machine Learning

This is a simple implementation of the classic Hangman game in C++. The program randomly selects a word from a predefined list of 2286 words and then allows the player to guess individual letters in an attempt to reveal the word. For every incorrect guess, a part of the hangman is drawn, and the player loses when the entire hangman is drawn.

Hangman Game

• Conducted statistical analysis and forecasting of solar energy data for two regions in Rajasthan using various time series models such as AR, MA, ARMA, ARIMA, and SARIMA. • Applied data preprocessing, descriptive statistics, distribution fitting, stationarity testing, and time series decomposition techniques to explore and visualize the data patterns and trends.

Statistical Analysis and Forecasting of Solar

A powerful and user-friendly tool leveraging LangChain framework and Gemini model to facilitate dynamic strategic assessments. Empower your decision-making with AI-driven insights for comprehensive analyses.

SWOT Analysis Generator

• Achieved 98.33% validation accuracy through extensive data augmentation and preprocessing techniques tailored to enhance model performance in wheat disease classification • Conducted a comparative evaluation of advanced deep learning architectures, including ResNet, EfficientNet, VGG, Inception, MobileNet, and DenseNet, to identify the most effective model for accurately classifying wheat rust diseases. • The developed model outperforms existing architectures, demonstrating notable improvements in precision, recall, and F1 scores critical for real-world applications.