Projects

Employee Performance Optimisation using ML in Google Cloud Platform | BigQuery, BigQueryML (May 2023 - July 2023)

• Leveraged GCP tools to visualize employee learning curves and develop data-driven employee assignment.
• Demonstrated the benefits of data-driven employee assignment & improved allocation of service employees.
• Utilized BigQuery, AutoML, BigQuery ML and Vertex AI for data exploration and training, Google data Studio (Looker) for visualizations.
• Used Python and Flask to create the User Interface (UI) and deployed it in Google App Engine (GAE).
• Employed Google Distance Matrix API for calculation of distances to client sites.

COVID-19 Data Analytics and Reporting using Azure | Data Factory, Databricks, PowerBI, SQL (Dec 2023)

• Built a solution architecture using Azure Data Factory, Azure Data Lake, Databricks, PoweBI and HDInsight.
• Integrated data from HTTP clients, Azure Blob Storage and Azure Data Lake Gen2 using Azure Data Factory.
• Created ADF pipelines to execute HDInsight activities and Databricks Notebook to carry out data transformations
• Created CI/CD pipeline for releasing Azure Data factory artefacts for deployment.

Validating Attribution Techniques using Deep Learning | CNN, XAI, PyTorch (May 2023 – Sept 2023)

• Aims to validate the robustness of saliency attribution techniques, providing explanations for model predictions.
• Randomly sampled 113 images from the training set and generated over 16,000 perturbed images, providing a robust foundation for conducting experiments and assessing model performance on unseen data.
• Implemented various attribution methods including GradCAM, SmoothgradCAM++, ScoreCAM and LayerCAM and tested it against ResNet18, ResNet50, VGG16, InceptionV3
• Used occlusion-based perturbation and adversarial noise assessment to reveal vulnerabilities and limitations in popular attribution methods, clarifying model interpretability, region sensitivity, and saliency map reliability.
• Leveraged advanced metrics including Pearson Correlation Coefficient, Earth Mover’s Distance, Area under Curve, Similarity, and Normalised Scanpath Saliency to provide comprehensive insights and rigorous evaluation.

Spam Detection using LSTM (Long Short-Term Memory) | RNN, TensorFlow (Jan 2024)

• Conducted data preprocessing, including cleaning and transformation, utilizing the Pandas library.
• Implemented LSTM architecture followed by a dense layer employing a sigmoid activation function.
• Utilized the Adam optimizer, binary crossentropy loss function, and accuracy as the evaluation metric.
• Visualized the loss and accuracy metrics using Matplotlib for both training and validation datasets.

Chatbot using Transformers in PyTorch | Transformers, PyTorch (Nov 2024)

• Aimed to develop a QA chatbot leveraging Transformers architecture in PyTorch.
• Preprocessed conversational datasets to train chatbot, ensuring data quality and relevance for effective learning.
• Evaluated the transformer model’s response generation capability through Greedy Decoding with batch size of 1, showcasing its effectiveness.

Movie Recommendation System using ML | Python, TKinter, SVD (Feb 2019 - June 2019)

• Spearheaded team of three, building a Collaborative based Movie Recommendation System leveraging Python.
• Implemented the SVD algorithm to mitigate MSE and RMSE and enhance system accuracy.
• Utilized Tkinter, as GUI, to design and develop an intuitive and visually compelling user interface.