My Works

Advanced AI chatbot integrating Flask, Pinecone, RAG, Langchain, OpenAI for dynamic, context-aware interactions.

Built an Ed-Tech platform for active retention learning using Flutter, Django, and Transformers at Inter-IIT.

Leveraged Variational Autoencoder and DCGANS to generate unique anime-style character faces with Pytorch.

Built a full-stack blog application with React Router, Redux, secure authentication, and file upload capabilities.

Utilizes vision transformers to estimate depth from single monocular images.

Generates multiple-choice questions using Langchain and OpenAI for educational purposes.

Classifies text into positive, neutral, and negative sentiments with probability predictions.

Demonstrated the working of Auto Code Generator using GPT-3 Neo, which generates code from snippets or commands. Built with the transformer model having 1.3 billion parameters.

Delivered a 2.5-hour live Python tutorial covering 30 key topics, hosted by Google Developer Students Club (GDSC). The video has over 3000+ views on YouTube.

Designed a deep imputation model using a sequence-to-sequence architecture to recover missing data sequences and implemented a sliding window algorithm for training samples.

Developed clustering and mathematical (Pulp) models for optimizing facility locations in warehouse management. Deployed code using Docker and increased efficiency by 2.1%, used service pods on an AWS Kubernetes cluster.

Worked on detecting YOLO models, i.e. yolov5-v7-v8 models, increased accuracy by 7.8% using FPN networks. Detected birds & calculated the interval, created dashboards on Streamlit, improved interference speed by 80fps.

Worked on scraping and generating a phone sales dataset with a MAPE score 0.029. Implemented LSTM, ARIMA and XGBoost model.

Evaluates Multimedia Automatic Misogyny Identification for Sem Eval Tasks.

Predicts movie preferences based on user behavior and movie attributes.

Detects traffic signs using Convolutional Neural Network (CNN) classifier.

Classifies emails as spam or ham using Natural Language Processing techniques.

Uses a neural network for binary classification of breast cancer cases.

Predicts prices using a neural network regression model.

Implements Support Vector Machines with grid search for optimal hyperparameter tuning.

Compares decision trees and random forests algorithms for predictive modeling.

Improves K-Nearest Neighbors algorithm performance through parameter tuning.

Applies logistic regression to predict survival outcomes on the Titanic dataset.

Predicts housing sales using linear regression techniques and performed EDA.