Projects
A selection of projects that I'm not too ashamed of
Fine-Tuning BERT for Sentence-Pair Classification
This repository contains a TensorFlow implementation that demonstrates how to fine-tune BERT-based models for the sentence-pair classification task. By leveraging the Hugging Face transformers and datasets libraries, this project enables researchers and developers to quickly experiment and evaluate various BERT models.
Telecom Churn Prediction and Driving factors
In this project we analyse customer-level data of a leading telecom firm, build predictive models to identify customers at high risk of churn and identify the main indicators of churn. We also provide recommendations to reduce churn.
Automatic Ticket Classification
This project uses NLP to streamline support ticket classification and resolution process. The automation of the customer support ticket system using NLP and ML techniques allows companies to efficiently identify and categorize support tickets, assign them to the relevant team for swift resolution, and continuously improve their products and services.
Image-to-Image Translation for MRI images Type
This project generates T1 MRI image from T2 and vice versa using CycleGAN(Generative adversarial networks) to reduce cost of performing MRI scans and delay in analysis.
CodePro Lead Scoring Pipeline design and implementation
In this project we design the pipeline structure for Lead Scoring for CodePro and implement the same using the AIrFlow to maximize profit through retraining pipeline and conducting development experimentation with MLFlow.
Disease Prediction using Patient Complaints (Named Entity Recognition using Custom CRF model)
This project predicts the disease using the patient’s complaints and share the treatment by performing Named entity Recognition using Custom CRF mode.
BoomBikes Bike Rent Volume Prediction and Business Recommendation
This project uses sales data to predict volume on a particular day and based on significant features we provided business recommendation.
Breast Cancer Classification (malignant tumour prediction)
This project classifies the tumours into malignant (cancerous) or benign(non-cancerous) on unbalanced class dataset. This also mention significant features and measure their significance.
Home Automation using Gesture Recognition using 3D-CNN and CNN-RNN
This project performs gesture recognition to perform automation of home appliances through recognition of the user's hand gestures.
Lending Club Loan default factor prediction and business recommendation
This project uses past loan data to isolate the driving factors of loan defaulting and credit loss. These factors are shared with business with recommendation to reduce credit loss and predict the future credit loss.
Melanoma Multiclass Classification using CNN
We designed system design draft for pipeline for company BeHealthy given requirements and budget for both open-source implementation and cloud implementation.
Pipeline System Design Draft for BeHealthy
We designed system design draft for pipeline for company BeHealthy given requirements and budget for both open-source implementation and cloud implementation.
Sentiment Analysis on IMDB Reviews
This project classifies reviews into Positive and Negative sentiment using the ML model trained on IMDB Reviews. This model uses Bag of words representation to process reviews to train the model.
Surprise House Price Prediction and Interpretation
This project predicts prices of houses for Surprise House company to maximize profit and explain significant factors for future planning. It also compares effect of L1 and L2 Regularization.
