Research Experience

Visual Intent Prediction and Dataset Development

Research Focus

Contributing to dataset development pipeline for visual intent prediction research, analyzing preexisting visual datasets and designing novel data collection methodologies aligned with research objectives.

Key Contributions

• Contributing to the dataset development pipeline, analyzing preexisting visual datasets and designing novel data collection methodologies
• Conducting comprehensive hyperparameter tuning and model optimization to enhance classification accuracy across diverse visual intent scenarios
• Working on visual intent prediction models using gaze data classification, implementing custom CNN and BiGRU architectures with multiple input channels based on EHTask framework
• Designed tailored train-validation-test splits to address data sparsity and categorical classification challenges

Technologies & Methods

Advanced Time Series Forecasting and Neural Architectures

Research Focus

Worked on advanced time series forecasting methods and led development of advanced neural architectures including LSTM RNNs and transformer models. Implemented and evaluated multiple AI approaches including zero-shot learning with LLMs.

Key Contributions

• Led development of advanced neural architectures including LSTM RNNs and transformer models
• Implemented and evaluated multiple AI approaches including zero-shot learning with LLMs
• Followed Agile methodology with daily standups and sprint planning
• Created detailed technical documentation and UML diagrams for architecture decisions
• Deployed models to cloud infrastructure using GitLab CI/CD pipelines
• Developed VayuBuddy, an air-quality prediction chatbot, using HuggingFace and LangChain implementations

Technologies & Methods